Analysis of the FM Radio Spectrum for Secondary Licensing of Low-Power Short-Range Cognitive Internet-Of-Things Devices Via Cognitive Radio

Analysis of the FM Radio Spectrum for Secondary Licensing of Low-Power Short-Range Cognitive Internet-of-Things Devices via Cognitive Radio

Derek Thomas Otermat

Bachelor of Science Electrical Engineering University of Florida 2008

Master of Science Electrical Engineering Florida Institute of Technology 2011

A dissertation submitted to the College of Electrical Engineering at Florida Institute of Technology in partial fulfillment of the requirements for the degree of:

Doctor of Philosophy in Electrical and Computer Engineering

Melbourne, Florida November, 2016

We the undersigned committee hereby recommend that the attached document be accepted as fulfilling in part the requirements for the degree of Doctor of Philosophy of Electrical Engineering.

“Analysis of the FM Radio Spectrum for Secondary Licensing of Short-Range Low-Power Cognitive Internet-of-Things Devices via Cognitive Radio,” a dissertation by Derek Thomas Otermat

______Ivica Kostanic, Ph.D. Associate Professor, Electrical and Computer Engineering Dissertation Advisor

______Carlos E. Otero, Ph.D. Associate Professor, Electrical and Computer Engineering

______Brian Lail, Ph.D. Associate Professor, Electrical and Computer Engineering

______Munevver Subasi, Ph.D. Associate Professor, Mathematical Sciences

______Samuel Kozaitis, Ph.D. Professor and Department Head, Electrical and Computer Engineering

Abstract

Title: Analysis of the FM Radio Spectrum for Secondary Licensing of Low-Power

Short-Range Cognitive Internet of Things Devices via Cognitive Radio

Author: Derek Thomas Otermat

Advisor: Ivica Kostanic, Ph.D.

The number of Internet of Things (IoT) devices is predicated to reach 200 billion by the year 2020. This rapid growth is introducing a new class of low-power short-range wireless devices that require the use of radio spectrum for the exchange of information.

To offset this extraneous demand for radio spectrum, the low-power short-range IoT devices need to utilize vacant spectrum through the use of Cognitive Radio (CR). The analysis presented in this dissertation indicates that the FM radio spectrum is underutilized in areas of the continental United States that have a population of 100,000 or less. These locations have vacant FM radio spectrum of at least 13 MHz with sufficient spectrum spacing between adjacent FM radio channels. The spectrum spacing provides the required bandwidth for data transmission and provides enough bandwidth to minimize interference introduced by neighboring predicted and unpredicted FM radio stations and other low-power short range IoT devices. To ensure that low-power short- range IoT devices maintain reliable communications vacant radio spectrum, such as the

FM radio spectrum in these areas, will need to be used through CR.

iii

Table of Contents

Abstract ...... iii List of Keywords ...... vii List of Figures ...... viii List of Tables...... ix List of Abbreviations...... x Dedication ...... xii Publications ...... xiii Chapter 1: Introduction ...... 1 1.1 Problem Statement ...... 1 1.2 Research Question ...... 1 1.3 Impact Statement ...... 2 1.4 Organization of the Dissertation ...... 2 Chapter 2: Literature Review ...... 4 2.1 Internet-of-Things ...... 4 2.1.1 Internet-of-Things Applications ...... 7 2.1.2 Internet-of-Things Enabling Technologies ...... 11 2.1.3 Cognitive Internet-of-Things ...... 13 2.2 Cognitive Radio ...... 17 2.2.1 Cognitive Radio Overview ...... 17 2.2.2 Spectrum Scarcity ...... 19 2.2.3 TV White Space ...... 23 Chapter 3: FM Radio ...... 28 3.1 FM Radio Zones ...... 28 3.2 FM Radio Station Classes ...... 28 3.3 FM Radio Protected Service Contours ...... 29 Chapter 4: FM Radio Spectrum Analysis Algorithm ...... 31 4.1 Step One: Generate State Station File ...... 32 4.2 Step Two: Calculate Station Distances to CUA ...... 33 4.3 Step Three: Remove Stations Outside Maximum Protected Service Contour 34 iv

4.4 Step Four: Remove Stations Outside Protected Service Contour ...... 34 4.5 Step Five: Calculate FM Stations Field Strength at CUA ...... 35 4.5.1 Propagation Model ...... 35 4.6 Step Six: Remove Stations with Field Strengths Less Than Protected Field Strengths ...... 36 Chapter 5: FM Radio Spectrum Measurements ...... 37 5.1 Measurement Locations ...... 37 5.2 Measurement Hardware ...... 38 5.2.1 Receiving Antenna ...... 38 5.2.2 Software Defined Radio ...... 40 5.3 Measurement Software ...... 42 5.4 Measurements ...... 44 5.4.1 Location 1 Measurements ...... 46 5.4.2 Location 2 Measurements ...... 49 5.4.3 Location 3 Measurements ...... 51 5.4.4 Location 4 Measurements ...... 53 5.4.5 Location 5 Measurements ...... 55 5.5 Comparison to Algorithm Results ...... 57 Chapter 6: United States FM Radio Maps ...... 59 6.1 Coordinates Under Analysis ...... 59 6.2 FM Radio Station Protected Coverage Map ...... 60 6.3 Unallocated FM Radio Spectrum Coverage Map ...... 66 6.4 Vacant FM Radio Spectrum ...... 69 6.5 CIoT Deployment Bitrates ...... 72 6.6 FM Radio Map Conclusions ...... 74 Chapter 7: CIoT Deployment in the FM Radio Spectrum ...... 77 7.1 Low-Power Short-Range CIoT Devices ...... 77 7.2 Spectrum Management: Access Control ...... 78 7.2.1 FM Radio Spectrum Geolocation Database ...... 79 7.2.2 Spectrum Sensing ...... 80 7.3 Spectrum Management: Spectrum Utilization ...... 82 7.3.1 Frequency Multiplexing ...... 82

7.3.2 Power Management ...... 83 Chapter 8: Summary and Conclusions ...... 85 8.1 Research Summary ...... 85 8.2 Conclusions ...... 87 References ...... 89

List of Keywords

1. Internet of Things

2. Cognitive Radio

3. FM Radio

4. Low-Power Short-Range Devices

5. Spectrum Scarcity

6. Spectrum Utilization

vii

List of Figures

Figure 2.2-1: “Table 5.1 Federal and Shared Bands Under Investigation for Shared Use” [11] ...... 20 Figure 2.2-2: Spectral Measurements: 0-90 MHz and 90-180 MHz [13] ...... 22 Figure 3.3-1: Six-Step Algorithm Flow Diagram [26]...... 32 Figure 5.1-1: The five locations where FM radio spectrum measurements were conducted ...... 37 Figure 5.2-1: Diamond D-130NJ antenna ...... 39 Figure 5.2-2: The RTL-SDR connected to the PE-SR405FL RF coaxial cable assembly ...... 40 Figure 5.2-3: Key components of the RTL-SDR [31] ...... 41 Figure 5.2-4: Internal architecture of the R820T/RTL2832U RTL-SDR [31] ...... 42 Figure 5.3-1: FM radio spectrum analyzer Simulink software diagram ...... 43 Figure 5.3-2: Spectrum analyzer plot of FM radio channel 222 at the first location ... 44 Figure 5.4-1: Setup of the measurement hardware at location 1...... 46 Figure 5.4-2: Location 1 predicted and measured received signal level strengths ...... 48 Figure 5.4-3: Setup of the measurement hardware at location 2...... 49 Figure 5.4-4: Location 2 predicted and measured received signal level strengths ...... 51 Figure 5.4-5: Location 3 predicted and measured received signal level strengths ...... 53 Figure 5.4-6: Setup of the measurement hardware at location 4...... 53 Figure 5.4-7: Location 4 predicted and measured received signal level strengths ...... 55 Figure 5.4-8: Location 5 predicted and measured received signal level strengths ...... 57 Figure 6.2-1: Continental United States FM radio station protected coverage map .... 60 Figure 6.3-1: The spectrum measurement of a HD FM radio station at location 2 ..... 66 Figure 6.3-2: Continental United States unallocated FM radio spectrum map ...... 68 Figure 7.3-1: Non-contiguous vacant FM radio channels at location 1 ...... 83

viii

List of Tables

Table 3.2-1: FM Station Classes and Service Contours [1] ...... 29 Table 5.4-1: Location 1 FM radio station information ...... 47 Table 5.4-2: Location 2 FM radio station information ...... 49 Table 5.4-3: Location 3 FM radio station information ...... 51 Table 5.4-4: Location 4 FM radio station information ...... 54 Table 5.4-5: Location 5 FM radio station information ...... 56 Table 6.2-1: Average number of FM radio stations for select locations ...... 61 Table 6.5-1: Potential CIoT deployment bitrates ...... 72

List of Abbreviations

ACI Adjacent Channel Interference ATSC Advanced Television Systems Committee BAS Building Automation System BER Bit Error Rate CIoT Cognitive Internet of Things CCI Co-Channel Interference COFDM Coded Orthogonal Frequency Division Multiplex CR Cognitive Radio DARPA Defense Advanced Research Projects Agency DVB-T Digital Video Broadcasting – Terrestrial EIRP Effective Isotropic Radiated Power ERP Effective Radiated Power EPC Electronic Product Code FCC Federal Communications Commission FIT Florida Institute of Technology FM Frequency Modulation GPS Global Positioning System H2H Human-to-Human H2T Human-to-Thing HAAT Height Above Average Terrain HD High Definition HL Hexagon Lattice HVAC Heating Ventilation and Air Conditioning IBOC In-Band On-Channel IF Intermediate Frequency IoT Internet-of-Things IP Internet Protocol IPv4 IP version 4 IPv6 IP version 6 ISM Industrial, Scientific, and Medical LPFM Low Power FM M2M Machine-to-Machine MIT Massachusetts Institute of Technology NC-OFDM Non-Contiguous Orthogonal Frequency Division Multiplexing (NC- OFDM) NTSC National Television System Committee OSI Open Systems Interconnection QPSK Quadrature Phase Shift Keying RFID Radio Frequency Identification RSL Received Signal Level SDR Software Defined Radio SIM Subscriber Identity Module SNR Signal-to-Noise Ratio x

SL Square Lattice T2T Thing-to-Thing TL Triangle Lattice TV Television TVWS TV White Space UHF Ultra High Frequency USB Universal Serial Bus VHF Very High Frequency WSD White Space Device WSN Wireless Sensor Network XG Next Generation Communications Program

Dedication

The pursuit of my doctorate degree, and all of the research and written work that it entailed, is dedicated to my wife Nicole, my son Carson, my daughter Norah, and my parents Thomas and Pamela. None of this would be possible if it were not for the love, encouragement, and support you all provided. Thank you.

I would also like to thank my doctoral advisor, Dr. Ivica Kostanic, and Dr. Carlos

E. Otero. Your help with my publications and research throughout my graduate career is greatly appreciated.

xii

Publications

D. T. Otermat, C. E. Otero and I. Kostanic, "Analysis of the FM Radio Spectrum for Internet of Things Opportunistic Access via Cognitive Radio," in 2015 IEEE 2nd World Forum on Internet of Things (WF-IoT), Milan, 2015. D. T. Otermat, I. Kostanic and C. E. Otero, "Analysis of the FM Radio Spectrum for Secondary Licensing of Low-Power Short-Range Cognitive Internet of Things Devices," IEEE Access, vol. PP, no. 99, pp. 1 - 1, 2016.

xiii

Chapter 1: Introduction

1.1 Problem Statement

The advent of the IoT has introduced a new class of low-power short-range wireless devices that require the use of radio spectrum. The increasing use of IoT devices, coupled with the rapid growth of wireless devices in general, is creating a problem by placing an overwhelming demand on the radio spectrum. This overwhelming demand is creating a shortage of radio spectrum required for new wireless devices. To combat the shortage of radio spectrum, IoT devices need to be able to identify vacant radio spectrum that can be used in an opportunistic manner through

Cognitive Radio (CR). To capitalize on the identified vacant radio spectrum IoT devices will be required to adjust their device parameters, such as frequency and transmit power, in order to optimize their throughput while minimizing interference to the primary spectrum license holder. Insufficient work has been done to identify efficient methodologies for low-power short-range IoT devices that opportunistically access vacant radio spectrum while mitigating interference to the primary spectrum license holder. [1]

1.2 Research Question

The principle research question addressed in this work is given as: How can vacant FM radio spectrum be used through CR to optimize the throughput of low-power short-range IoT systems while minimizing interference to the FM radio stations

(primary users)? 1

1.3 Impact Statement

The proposed research will have a significant impact on the deployment of low- power short-range IoT devices. It considers methods for providing access to underutilized spectrum through the novel use of the vacant FM radio spectrum via CR.

As more IoT devices are deployed, the available radio spectrum will become congested and access will be scarce. Therefore, to help ensure that IoT devices have access to radio spectrum for communications, it is imperative that these IoT devices use vacant radio spectrum such as the vacant FM radio spectrum. Low-power short-range IoT devices are becoming an integral part of industries such as the consumer, manufacturing, and medical industries. Poor communications, caused by inadequate access to radio spectrum, will cause loss of data and could lead to extended periods where there is a complete loss of communications. Ensuring communications by providing sufficient access to radio spectrum, specifically in the medical industry, could save lives by ensuring critical data is not lost between medical IoT devices. [1]

1.4 Organization of the Dissertation

This dissertation is organized as follows: Chapter 2 provides a review of the current

IoT and CR literature. It also identifies possibility of FM radio spectrum utilization for low-power short-range Cognitive Internet of Things (CIoT) devices. Chapter 3 provides an overview of the characteristics of FM radio that are pertinent to the research in this dissertation. Chapter 4 describes the algorithm that is used to quantify the unallocated

FM radio spectrum for a given geographical location. Chapter 5 presents the spectral measurements that were conducted in the FM radio spectrum in order to validate the

algorithm discussed in chapter 4; Chapter 6 showcases the maps that quantify the number of FM radio stations and the amount of unallocated FM radio spectrum and discusses the achievable CIoT bit rates across the continental United States. Chapter 7 presents methodologies that low-power short-range CIoT devices may employ to use the vacant FM radio spectrum through CR; Chapter 8 provides a research summary and concludes the results.

Chapter 2: Literature Review

To formulate a sound understanding of the literature pertaining to this dissertation’s research question, many journal and conference papers were reviewed and analyzed.

The research question can be divided into two primary areas, IoT and CR. In addition, the research presented in this dissertation relies heavily on the fundamental details of

FM radio and, therefore, an overview of it becomes essential and will be discussed following the literature review. To the author’s knowledge, other than this dissertation’s published research [1], there are no publications pertaining to the use of the FM radio spectrum, via CR, for IoT applications. Furthermore, there is no published work that analyzes the FM radio spectrum in order to fully understand its vast potential and capacity for CR applications.

2.1 Internet-of-Things

The IoT is growing at an extreme rate and at the current rate of deployment, eventually trillions of IoT devices will be on the Internet [2]. Consequently, this rate of deployment places an overwhelming demand on the radio spectrum. To combat this demand, it is imperative that CR be implemented in IoT devices to ensure the reliable transmission of information. Furthermore, various portions of the spectrum, such as the

FM radio spectrum, will need to be analyzed for the viability of secondary licensing through CR.

IoT devices are typically deployed using one of three widely used deployment patterns; Square Lattice (SL), Triangle Lattice (TL), or Hexagon Lattice (HL) [3]. The performance of these deployment patterns, when they are used through CR, is studied

in [3]. The rapid rate of IoT deployment is causing dense device distributions for these patterns [3]. Dense distributions are causing transmission conflicts and poor device energy efficiency [3]. To alleviate the dense distributions, CR can be implemented in a spectrum band that can accommodate the growing number of IoT devices.

The current primary mode of Internet communications is Human-to-Human

(H2H) [4]. H2H communications are communications that occur between devices that require human interaction. Examples of H2H communications are e-mail and social media exchanges via a personal computer or voice conversations and text messaging via a smart phone. With the advent of IoT, the modes of Internet communications will expand to include Human-to-Thing (H2T) and Thing-to-Thing (T2T) [4]. T2T communications are also known as Machine-to-Machine (M2M) communications and are one of the key fundamental building blocks of IoT.

The IoT is comprised of three visions: a things oriented vision, an Internet oriented vision, and a semantics oriented vision [5]. The things oriented vision is dependent on the emerging wireless technologies that enable new IoT devices. As IoT devices become interleaved in our everyday activities it is imperative that CR is integrated into IoT devices to mitigate spectrum congestion and to ensure reliable communications. The Internet oriented vision is based on the need to have the IoT devices connected to each other [5]. The IoT devices all need to have compatible network layers. One of prominent network layer protocols being used on the Internet is

Internet Protocol (IP) [5]. The current version of IP, IP version 4 (IPv4), uses 32-bit addresses that are divided into four byte chunks. Since the address length is 32 bits, the total number of addresses available to devices is 232 = 4,294,967,296. Though IPv4 has

approximately 4.3 billion addresses, it is quickly approaching maximum capacity. To combat this, IP version 6 (IPv6) is being deployed. IPv6 uses 128-bit addresses and has a total of 2128 = 3.4 x 1038 addresses available to devices. This means that there are approximately 8 x 1023 more addresses available in IPv6 than there are in IPv4. The semantic oriented vision addresses data processing. The semantic oriented vision is based on the fact that the number of IoT devices that will be generating data will be huge [5]. It is important to have efficient data processing algorithms and abundant data storage to store all of the information generated by IoT devices. Having consistent data formats will be key to developing efficient data processing algorithms. The IoT paradigm is dependent on the successful execution of all three IoT visions.

The IoT is a vision where “things” or “objects” become part of the Internet [6].

By being part of the Internet, the IoT M2M devices are uniquely identified and accessible to other M2M devices [6]. These M2M devices, such as sensors, actuators, etc., are able to interact and cooperate with one another to reach a common goal [7]. IoT itself is not a new type of technology rather, a new paradigm that is made possible by evolving enabling technologies. This new paradigm, initiated by the concept of smart environments, paves the way for the deployment of numerous applications and provides a significant impact on many fields and the future of everyday life [7].

As the vision of IoT continues to mature at a rapid pace, its importance is being globally recognized. The U.S. National Intelligence Council predicts that IoT is among six technologies that will impact U.S. national power by 2025 [7]. The European

Commission has realized the potential of IoT and therefore, is funding several projects within the VII Framework Program [7]. In 2010 China announced that they are

designating Wuxi City the National Innovation Demonstration Zone of IoT to promote

IoT related research and development [7].

The scope of this literature review, as it pertains to IoT, is to first provide an overview of the IoT applications. Second, the technologies that enable IoT devices will be discussed. The enabling technologies are discussed to show the dependence of IoT devices on wireless technologies. This dependence on wireless technologies will eventually place an overwhelming demand on the radio spectrum which will ultimately lead to spectrum scarcity. To combat spectrum scarcity CR technology will need to be implemented in IoT devices. Third, CR technology, and the role it currently has in IoT devices, will be discussed.

2.1.1 Internet-of-Things Applications

The rapid employment of IoT devices is a direct result of the large number of possible application domains. The authors of [8] divide the IoT application domain into six areas:

1. Smart Home and Smart Building

2. Healthcare

3. Smart Business

4. Utilities

5. Traffic Monitoring

6. Environmental Monitoring

2.1.1.1 Smart Home and Smart Building

Household devices have embraced technologies which enable them to become interactive with humans and other devices. Household devices such as air conditioning 7

and heating systems washing machines, refrigerators, etc. can provide H2T and M2M interaction. For example, these devices provide H2T communications by alerting humans via wireless enabled devices, such as smart phones or tablets, that maintenance and servicing of the respective household device is required. In addition, humans can initiate diagnostics when trying to troubleshoot a problem with the household device.

These household devices can communicate M2M as part of the IoT to optimize energy consumption and lower energy bills [8].

IoT devices can be configured to serve as a Building Automation System (BAS).

BAS’s are designed to optimize cost effectiveness while maintaining the system’s overall efficiency [8]. BAS’s primary focus is on energy savings via regulating Heating

Ventilation and Air Conditioning (HVAC) systems and lighting [8]. In addition to lowering energy expenses, the BAS also provides management of security and safety devices [8].

2.1.1.2 Healthcare

IoT devices can be deployed in the healthcare field to benefit both the health care providers and patients alike. IoT devices can benefit doctors and nurses by providing mechanisms to remotely monitory a patient’s vital signs such as blood pressure, heart rate, oxygen levels, temperature [8]. This will allow doctors and nurses to efficiently monitor more patients while ensuring a high level of care. In addition to facilitating efficient patient monitoring, IoT devices can alert the doctors and nurses when a patient’s vital signs exceed a certain threshold, which could indicate an emergency situation where time is of the essence.

Many of the benefits that IoT devices provide to the doctors and nurses, also hold true for the patients receiving the medical care. Patients benefit from having their vital signs monitored because if a vital measurement exceeds a certain value, it will alarm the health care providers to a potential emergency situation. This could potentially be lifesaving. Having vital signs monitored and analyzed over a time period provides the health care providers with data that may be used to adjust a patient’s treatment or medication regimen to improve a patient’s quality of life. In addition to improving a patient’s quality of life, alerting health care providers of abnormalities in a patients vital measurements could reduce health care costs by facilitating early intervention and treatment [5].

2.1.1.3 Smart Business

IoT devices are being implemented into many different types of businesses and industries. IoT devices, enabled by Radio Frequency Identification (RFID) technology, provide a mechanism for monitoring and managing the movement of assets [8] in the logistics and inventory management industries. RFID sensors are affixed to a tracked asset and can either be passively or actively read by RFID sensing devices.

IoT devices can be used in the agricultural industry for monitoring soil properties. This enables educated decision making in the production of natural food ingredients, prevents loss of crops [5] and increases profit. Water conservation is also an industry that can benefit from the IoT. In regions where droughts are often a concern,

IoT devices may be used to mitigate the wasting of water and help with water conservation [5].

2.1.1.4 Utilities

The primary motivation behind incorporating IoT devices into the utilities market is optimizing efficiency and cost savings. Utility companies that provide energy, water, natural gas, propane, etc. have to charge extra costs to cover the expenses that are incurred by manually reading and analyzing consumer usage data [8]. Remotely measuring and monitoring consumer usage data lowers costs by reducing the amount of time spent by an employee physically measuring and analyzing the usage of each individual household.

IoT devices can increase resource conservation and lower costs by allowing companies to remotely monitor the homeowner’s utility consumption. Companies can monitor household utility consumption and inform the homeowners when measured data indicates the over usage of resources. This will not only conserve resources but it will also reduce costs. The homeowner can immediately reduce consumption, rather than after they receive a large bill. Over consuming resources can also be an indication of a faulty system at the homeowner’s residence that could require maintenance or replacement. Alerting the homeowners could facilitate early detection of an issue. Early detection of an issue could result in a simple repair, mitigating a major repair or the replacement of an expensive device.

2.1.1.5 Traffic Monitoring

IoT devices used for traffic monitoring applications are primarily enabled by

Wireless Sensor Network (WSN) technology. These IoT devices provide travelers with an update to traffic, air pollution, and noise emission conditions to aid in choosing the most desired route [8]. An algorithm using data collected from IoT devices can be

developed to automatically update the GPS selected route to mitigate road construction and hazards to reduce travel time [8]. This reduces fuel consumption and economic losses [5]. IoT devices can also be installed onto public transportation vehicles such as buses, trains, and taxis, to inform the travelers of the vehicle’s estimated arrival times

[8].

2.1.1.6 Environmental Monitoring

IoT devices can be used to measure, monitor, and analyze naturally occurring phenomena like temperature, humidity, pressure, rainfall, water levels, wind speeds, etc.

[8, 5]. An important application of this is monitoring the drinking water to ensure the quality of the water is at a level that is safe for human consumption [8].

2.1.2 Internet-of-Things Enabling Technologies

A common trait amongst the IoT literature is the dependence on enabling technologies. The specific type of enabling technology is dependent upon which layer of the IoT architecture is being considered. The authors of [4] propose a variant of the

Open Systems Interconnection (OSI) model for the IoT architecture. The first layer is split between Existed Alone Application System Layer and Edge Technology

Layer/Access Layer [4]. The second, third, fourth, and fifth layers are Backbone

Network Layer, Coordination Layer, Middleware Layer, and Application Layer, respectively [4]. Each of these different layers possess its own enabling technology. The focus of the research discussed in this dissertation is the enabling technology that comprises the Physical Layer, or the Edge Technology Layer.

With respect to the IoT Physical Layer, the enabling technologies are predominately wireless architectures, which makes communications between IoT 11

devices possible. Currently, IoT devices are enabled by primarily two types of wireless device classes: RFID and WSNs. One could reasonably argue that RFID devices are primitive in capability however, it is important to identify their crucial role in the IoT.

RFID devices are one of the pivotal enabling technologies of IoT [4]. The fundamental concept behind RFID is identifying any object, or IoT device, using specifications of Electronic Product Code (EPC) [5]. The goal of EPC is supporting the use of RFID devices in conjunction with the Internet to promote their use in IoT and to support the growth of smart industry [8]. EPC was developed by Auto-ID from the

Massachusetts Institute of Technology (MIT) with the purpose of sharing data in real- time through the use of RFID technology [8]. RFID technology is comprised of RFID tags, RFID tag readers, antennas that are integrated into the RFID tags, and database management software [8]. Though RFID devices were the pioneering technology during the beginning of IoT, it is now primitive in comparison to new technologies. That is not to say that RFID devices do not have a relevant purpose in the IoT however, complex

M2M communications have been facilitated using WSNs.

Technological advances in the WSNs, specifically in the hardware domain, have produced cost-effective devices that can be used in IoT applications [5]. A WSN device, which is used as an IoT device, has interfaces to sensors, computing and processing units, transceivers, and power sources [5]. Unlike RFID devices, WSN devices have the ability to interact intelligently, in a real-time sense, to their environment. RFID devices typically produce information that is programmed in their hardware a priori. This difference is what makes WSN devices a front runner for the future of IoT devices.

One commonality between RFID, WSN, and any wireless IoT device is the dependence on radio spectrum for successful communications. The ubiquitous growth of IoT devices, employing wireless technologies as their primary enabling technology, is placing an overwhelming demand on the radio spectrum. This overwhelming demand, primarily in unlicensed frequency bands such as the Industrial, Scientific, and Medical

(ISM) band, coupled with the current frequency licensing and access policies are causing the spectrum to become a scarce commodity. To combat spectrum scarcity, opportunistic access to the radio spectrum must be exercised. The use of CR technology will enable IoT devices to opportunistically access vacant radio spectrum. Limited research of CR IoT technology has been conducted.

2.1.3 Cognitive Internet-of-Things

Cognitive M2M communications is a vastly unexplored field where limited research has been conducted [9]. M2M devices are the fundamental building blocks of the IoT. Therefore, from this point forward in this dissertation, the terms “M2M” and

“IoT” will be used interchangeably. The main challenge in IoT communications is allocating spectrum for the ever increasing number of IoT devices. According to

Ericsson, 50 billion connected devices will exist by 2020 [9]. This many devices competing for wireless spectrum will cause severe congestion. Mitigating spectrum congestion is the primary reason for incorporating CR in IoT [10]. The Cognitive IoT

(CIoT) section of this literature review will discuss the research that has been conducted on the types of CIoT devices, the challenges that IoT devices are faced with that can be alleviated through the use of CR, and coexistence amongst CIoT devices.

IoT devices can be divided into two types: capillary IoT devices and cellular IoT devices [9]. Cellular IoT devices are equipped with embedded Subscriber Identity

Module (SIM) cards and are able to communicate with other cellular IoT devices via cellular networks [9]. Since cellular IoT devices communicate via cellular networks, they have the capability to communicate over long distances. The research presented in this dissertation is limited to short-range IoT devices therefore, cellular devices will not be discussed further. Capillary IoT devices communicate with each other via short-range communication technologies such as Zigbee and Wi-Fi [9]. If capillary IoT devices need to communicate with other IoT devices that are out-of-range, gateway devices are used

[9]. Capillary IoT devices are generally required to be energy efficient and have a high reliability [9].

The authors of [9] identify three major challenges that IoT devices are faced with as a result of the rapid growth of IoT. The three challenges are spectrum scarcity, interference, and coverage issues [9]. The first challenge, spectrum scarcity, is a common theme in this dissertation and is discussed throughout. Spectrum scarcity is the motivating driver behind implementation of CR. The large-scale deployment of IoT devices will cause congestion and contention in the radio spectrum, making it a scarce resource. Using CR in IoT devices will help alleviate the spectrum congestion by offering efficient use of underutilized radio spectrum [9]. The research conducted herein analyzes the FM radio spectrum to determine its suitability for support of CIoT devices.

The second challenge, interference, is related to spectrum scarcity. Traditional wireless technologies force IoT devices to share unlicensed frequency bands, such as the ISM band, amongst numerous unlicensed wireless devices. The resulting spectrum

scarcity will inherently cause interference issues between IoT devices and non-IoT devices. Using CR in IoT devices will allow communications in underutilized spectral bands. One underutilized spectral band that is getting a lot of attention in academic research is the former analog Television (TV) band. The switch from analog to digital

TV transmission has left the legacy analog TV channels vacant. The use of these vacant

TV channels, known as TV White Space (TVWS), is being investigated for secondary use via CR [9].

The third major challenge that IoT devices are subjected to is wireless communications coverage. In many applications, IoT devices are forced to operate in various physical locations [9]. Therefore, wireless propagation in those different areas is not always guaranteed [9]. This is especially true in unlicensed frequency bands such as ISM [9]. CIoT devices have the flexibility to use spectral bands, such as TVWS, that provide better propagation characteristics [9].

The last section of the CIoT portion of the literature review addresses coexistence between CIoT devices. Multiple CIoT devices will most likely co-exist in a small area.

For example, CIoT devices could co-exist in a household for applications such as multimedia distribution and networked smart appliances [10]. Co-existence between

CIoT devices in unlicensed spectral bands, such as the ISM band, is more easily obtained since regulators have imposed low Effective Isotropic Radiated Power (EIRP) limits, 100 mW in Europe and up to 1 W in the United States [10]. Co-existence in licensed spectral bands that have been provisioned for secondary licensing through CR such as the TVWS, typically do not have EIRP limitations and can consist of many different CR technologies. A mixture of wireless technologies with different

transmission power levels, distances, network architecture, and terminal capabilities makes co-existence between CIoT devices challenging [10]. To mitigate co-existence issues between CIoT devices, proper power management and frequency reuse schemes must be utilized. In addition, the CIoT devices must transmit with optimal transmission characteristics to achieve the desired throughput, while maximizing the number of CIoT devices that can co-exist. One important aspect that the authors of [10] failed to address is the co-existence of CIoT devices with the licensed frequency holder, known in CR vernacular as the primary user. It is imperative that all CIoT devices mitigate interference to the primary user.

This portion of the literature review provided an overview of the IoT applications, the technologies that enable IoT devices, and the role CR currently has in IoT devices.

The rapid expansion of IoT devices will cause the unlicensed spectrum bands to become congested and scarce. Therefore, it is vital that CR be incorporated into IoT devices.

Currently, the only frequency band being investigated for CIoT is the TVWS. This is expected, as it is the spectral band that has been subjected to the most research.

However, it is also important to investigate other frequency bands because as the use of the TVWS increases, it will become congested by both CR and primary users. The research contained in this dissertation investigates the FM radio spectrum to determine if it can be utilized by CR for CIoT devices.

2.2 Cognitive Radio

2.2.1 Cognitive Radio Overview

One of the most revolutionary applications of CR is addressing spectrum scarcity in wireless communications [10]. The spectrum is scarce primarily because of the way it is licensed. This is evident through the spectrum occupancy studies documented in

[11, 12, 13]. Both [12, 13] identified spectrum occupancy of approximately 20%, indicating that the spectrum is 80% underutilized. If the spectrum were to be shared amongst multiple licensees, a much higher utilization could be achieved. CR provides the technical framework for spectrum sharing of the underutilized spectrum. Harnessing the underutilized spectrum for use by CIoT devices will be key to the future success of

IoT networks.

The author of [14] provides an excellent description of CR and the tasks that are required for the success of CR. The author defines CR as “an intelligent wireless communication system that is aware of its surrounding environment (i.e. outside world), and uses the methodology of understanding-by-building from the environment and adapt its internal states to statistical variations in the incoming RF stimuli by making corresponding changes in certain operating parameters (e.g., transmit-power, carrier frequency, and modulation strategy) in real-time, with two primary objectives in mind:

(1) highly reliable communications whenever and wherever needed and (2) efficient utilization of radio spectrum” [14]. This thorough definition identifies three operating parameters that are varied in real-time to ensure proper functionality of a CR, transmit power, carrier frequency, and modulation scheme [14]. These three operating parameters are key to the success of the research presented herein. 17

The first operating parameter, transmit power, will be addressed in Chapter 7 of this dissertation. Power management will be critical for two reasons: ensuring enough

Signal-to-Noise Ratio (SNR) at the CR receiver and limiting interference to the primary and secondary, or CR, users. The first reason is related to the third operating parameter, the modulation scheme. For a specific task or application a minimum Bit Error Rate

(BER) will be defined. The CR transmitter’s transmit power will be adjusted to ensure sufficient SNR at the CR receiver. If the transmit power is required to be adjusted to an unacceptable level, the modulation scheme may have to be changed. There will be a continual balance between the transmit power and modulation scheme. Also, the BER can be adjusted while maintaining a constant transmit power. The second reason, limiting interference to the primary and secondary users, will not be as difficult of a challenge with respect to the primary user’s interference as it is when CR is used in other frequency bands. This is because the FM radio station coverage map is consistent over short distances. For example, if the vacant FM radio spectrum were used to connect a printer to a desktop computer inside of a home or office, interference to the primary users i.e. the FM radio stations should not be a major concern because the FM radio station coverage map is unlikely to change inside a home or office. In this situation, the concern would be causing interference to other CR users utilizing the vacant FM radio spectrum. Power management can be employed to ensure multiple CR users can utilized concurrent portions of the vacant FM radio spectrum.

The second operating parameter is carrier frequency. As the FM radio station coverage map varies, the channels that service a particular location vary. Each of the channels has an associated center frequency with 200 kHz of bandwidth. Devices using

vacant FM radio spectrum will be required to utilize the vacant radio station channels for data transmission. If the device is stationary, then the FM radio station coverage map should not vary, unless new stations are added, and the vacant radio station search will be trivial. If the device is deployed in mobile applications, the vacant radio station search will not be as trivial. In this situation, the device will need to access the FMWS database.

In addition, the device will be equipped with the ability to sense the spectrum in real- time to determine the vacant radio channels in the event that the device cannot access a database containing the vacant FM radio spectrum.

The third operating parameter, modulation scheme, was discussed in conjunction with the first operating parameter, transmit power. The modulation scheme will adaptively selected to maximize the symbol rate while maintaining a SNR that does not force the previously discussed transmit power limitations.

2.2.2 Spectrum Scarcity

To show how vacant FM radio spectrum be used through CR to optimize the throughput of low-power short-range IoT systems while minimizing interference to the

FM radio stations, it is imperative to identify what problem is trying to be solved.

Review of the spectrum scarcity research was conducted to prove that spectrum scarcity is indeed a pseudo situation. It is pseudo in the sense that the spectrum remains idle or vacant the majority of the time however, it cannot be utilized by anyone but the primary license holder. The first piece of spectrum scarcity literature that was analyzed was a report presented to the president of the United Sates entitled “Realizing the Full

Potential of Government-Held Spectrum to Spur Economic Growth” [11]. This report identifies major flaws in the current spectrum licensing scheme which is producing 19

severe underutilization of the spectrum. This report recommends the immediate clearing of 1000 MHz of spectrum for wideband communications [11]. Figure 2.2.1-1 identifies the recommended federally held spectrum to be cleared.

Figure 2.2-1: “Table 5.1 Federal and Shared Bands Under Investigation for Shared Use” [11]

The frequency bands identified were chosen because they possess characteristics that make them highly favorable for use [11]. For example, the 2.4 GHz ISM band used by

IEEE 802.11 Wi-Fi is becoming very congested and the 5.8 GHz ISM band has poor propagation characteristics [11]. Instead, the 2.7 – 3.6 GHz frequency band offers less congestion than the 2.4 GHz band and better propagation characteristics than the 5.8

GHz band [11]. This is just one example of the benefits a shared spectrum access scheme provides. This report recommends the use of a geo-location database that is currently being utilized by TVWS standards. To determine if a particular band is being occupied, a device queries a geo-location database to determine if a primary user is registered in that band for that respective geographical location. This mode of operation is not reliant

on the traditional CR methodology, where the spectral band is monitored real-time for primary user activity. One drawback that exists with having a geo-location database only access scheme is the reliance on consistent connectivity to the geo-location database. If a device is not able to query the geo-location database, traditional CR technologies i.e. spectrum sensing techniques should be able to be employed to ensure consistent spectral access. For spectral environments that can change rapidly, as is the case for ISM bands, it is imperative to have the ability to sense if a desired spectral band is occupied. Relying solely on a geo-location database does not provide this ability.

Since occupancy of FM radio channels does not change rapidly, reliance on a geo- location database should suffice. However, in order to access the geo-location database, the CR would have to have access to the Internet. Depending on the application, the CR may not have access to the geo-location database via the Internet. To remain versatile, the CR should also have the ability to real-time sense the surrounding FM radio channels to determine the existence of the primary user.

The second piece of spectrum scarcity literature analyzed the spectral occupancy of the Very High Frequency (VHF) frequency band, 30 – 300 MHz, in the large U.S. city of Columbus, Ohio [13]. The objective of this study was to quantify the spectral occupancy of the VHF frequency band to show the potential benefit of CR technology.

In addition to the measurements taken in a large city, measurements were conducted to capture the spectral occupancy of a rural environment. This study estimates that approximately 80% of the 30 - 60 MHz band is essentially free of signals to within about

7 dB of the environmental noise limit at 30 kHz resolution [13]. Two spectral bands that the study does not analyze or measure are the TV and FM radio bands. The study states

that there are large portions of the spectrum which are occupied by very strong, relatively broadband signals i.e. TV and FM radio and there isn't any reason to include these in the spectral occupancy calculations [13]. However, the analyses proposed herein will prove that the FM radio spectrum is underutilized and CR can be utilized to harness the underutilized FM radio spectrum. Figure 2.2.-2 shows the measured spectral occupancy from 0 to 90 MHz on the left plot and the 90 to 180 MHz on the right plot.

It is clearly evident that the 30 to 60 MHz band is extremely underutilized. Annotated in Figure 2.2-2 is the FM radio spectrum that is sparsely utilized, primarily in the right plot. The research presented herein will make use of the underutilized FM radio spectrum for secondary licensing through CR. In Figure 2.2-2 the red trace represents the maximum measured values, the blue trace represents the mean measured values, and the green trace represents the mean measured values with a matched load replacing the antenna [13].

Figure 2.2-2: Spectral Measurements: 0-90 MHz and 90-180 MHz [13] The last piece of spectrum scarcity literature that was reviewed was a spectrum occupancy study performed in Chicago, Illinois [12]. The preceding piece of literature analyzed the spectrum occupancy of the VHF band. This paper analyzed a wider band,

30 MHz to 3 GHz over a 48-hour time period [12]. Like [13], this paper substantiates the fact that the spectrum is widely underutilized. This paper measured the spectral occupancy to be 17.4%. This corresponds to the spectrum being 82.6% underutilized.

This paper analyzed a wide range of frequencies, however, the FM radio spectrum was not analyzed. The results presented in this and the preceding paper both show that the spectrum remains approximately 80% underutilized. To mitigate the pseudo-scarce spectrum paradigm, CR can be employed to make efficient use of the identified underutilized spectrum.

2.2.3 TV White Space

The preceding two sections discussed papers that identified the problem of spectrum scarcity, which consequently leads to spectrum underutilization because of the way the spectrum is licensed, and a technical means to efficiently use the underutilized spectrum, CR. CR provides a means for using vacant spectrum, or white space, but it is not biased towards any particular frequency bands. One frequency band that is getting a lot of attention is the TVWS band.

The authors of [15] discuss the origin of the TVWS development, the regulatory status, and the standards that have been developed to use this white space. The authors define TVWS as “TV channels that are not used by any licensed services at a particular location and a particular time [15]”. Like TVWS, the vacant FM radio spectrum is dependent on location. FM channels that are occupied by FM radio stations in one city may be vacant 100 miles away. However, unlike TVWS, the vacant FM radio spectrum will not fluctuate much with time. The TVWS can fluctuate over time because broadcast microphones can legally occupy the spectrum. If a FM channel is allocated to a FM 23

radio station, the only time the vacancy would change is if the FM radio station went off the air and discontinued service to that particular area.

The TVWS rules were drafted by the Federal Communications Commission

(FCC) after the commission saw the positive attributes of Defense Advanced Research

Projects Agency’s (DARPA) Next Generation Communications Program (XG) radio

[15]. DARPA’s XG radio was developed for the United States military with the goal of providing troops with radios that could exploit vacant spectrum [15]. The FCC has levied TVWS regulations by amending Parts 0 and 15 of Title 47 of the Code of Federal

Regulations [15]. The FCC regulations protect primary users such as primary TV broadcast stations and wireless microphones [15].

In TV broadcasting two frequency bands, VHF and Ultra High Frequency (UHF), are allocated for use [15]. The FCC has established TVWS in both VHF and UHF bands

[15]. In the United Kingdom, only the UHF is allowed to facilitate TVWS [15]. In the

United States the TV channel bandwidth is 6 MHz and in the United Kingdom the channel bandwidth is 8 MHz [15]. The FCC has defined three regulations for protecting the primary users of the TV channels from secondary users [15]:

1. Utilization of a TVWS database that contains the vacant TV channels for a

specific area

2. Adherence to transmit power limitations

3. Ability to perform real-time TV channel spectrum sensing

The FCC has defined four classes of TVWS secondary user devices: fixed, portable mode II, portable mode I, and sensing only [15]. Fixed devices reside at a registered location, uses an outdoor antenna(s) and may transmit a maximum of 1 W

into one or more of the 6 MHz TV channels [15]. Antenna gains up to 6 dBi are permitted, resulting in an EIRP up to 4 W [15]. Access to the TVWS database is required but real-time spectrum sensing is not [15]. These types of devices cannot be used in a vacant channel that is adjacent to an occupied TV channel [15].

Portable mode II and I devices are lower in power and are restricted to operations in frequency bands 512 – 608 MHz (TV channels 21 – 36) and 614 – 698 MHz (TV channels 38 – 51) [15]. These devices are allowed an EIRP up to 100 mW (20 dBm)

[15]. If these types of devices are used in a vacant channel that is adjacent to an occupied

TV channel, the maximum EIRP values are reduced from 100 mW to 40 mW [15].

Mode II devices require access to the TVWS database however, mode I devices do not

[15]. Neither mode II nor I devices require real-time spectrum sensing capabilities [15].

Sensing devices are allowed an EIRP up to 50 mW (17 dBm) [15]. These devices must have the real-time sensing capability to sense Advanced Television Systems

Committee (ATSC) digital TV signals and National Television System Committee

(NTSC) analog TV signals at -114 dBm per 6 MHz [15]. In addition, wireless microphone signals at -107 dBm per 200 kHz are also required to be detected [15].

Laboratory and field tests administered by the FCC are mandatory for the approval of sensing only devices [15].

Similar classifications for CIoT devices using the vacant FM radio spectrum could be deployed. CIoT device classes could be divided into stationary and mobile devices.

EIRP limitations have not been established, but the limitations would be chosen to mitigate interference to adjacent FM radio stations and other CIoT secondary users.

Mobile devices would require access to a vacant FM radio spectrum database and real-

time spectrum sensing capability. Stationary devices would only require access to a vacant FM radio spectrum database.

To quantify the amount of vacant spectrum, or TVWS, available for CR users, various studies have been performed. The first study, focused in southern Greece, quantified the amount of TVWS available using a geolocation-based approach [16]. The second study, which focused on the continental United States, quantified the amount of

TVWS using the FCC’s database of television transmitters coupled with wireless propagation models [17]. A third study was performed that analyzed the available

TVWS with respect to the operational requirements of White Space Devices (WSD)

[18]. WSD are CR devices that use the TVWS. The authors of [18] analyzed the availability based on RF characteristics such as the co-channel interference (CCI), adjacent channel interference (ACI), and WSD bandwidth requirements. Analyzing the available white space based on the WSD requirements is a novel approach that can be used during the evaluation of the white space required for low-power short-range CIoT devices. Though these studies do not quantify the vacant FM radio spectrum however, they provide excellent methodologies that may be used during future quantification of the vacant FM radio spectrum. In addition to quantifying the available TVWS, it is vital that the coexistence between CR devices is addressed. The authors of [19] addressed coexistence by presenting an independent framework that uses centralized and distributed architectures. Using an effective coexistence architecture will be key to the success of low-power short-range CIoT devices that use CR technology, by efficiently sharing frequency allocations to maximize throughput.

The literature review highlighted the need for scalable solutions that can help alleviate the spectrum scarcity problem faced by future IoT devices. Therefore, there is a need for a formal methodology to determine suitability of vacant FM radio spectrum to support IoT systems and device communication via CR (on a specific geographical area). This methodology will enable efficient deployment of IoT devices and support the future vision of self-managed IoT devices.

Chapter 3: FM Radio

The FM radio descriptions in this chapter, and the subsequent subchapters, are transcribed from [1]. FM radio spectrum consists of 100 channels that occupy the 88

MHz to 108 MHz spectral band [20]. These 100 channels are designated numbers 201 through 300 [20]. Channel 201 has a center frequency of 88.1 MHz and channel 300 has a center frequency of 107.9 MHz [20]. Each FM radio channel is allocated 200 kHz of usable bandwidth. The FM radio channels can be utilized for commercial, noncommercial, and educational broadcasting. Channels 201 through 220 are designated for non-commercial or educational use [21].

3.1 FM Radio Zones

In the United States, the geographical FM radio broadcast map is broken down into three zones: Zone I, Zone I-A and Zone II. Zone I and I-A is comprised of the states

CA (south of 40° latitude), CT, DC, DE, IL, IN, MA, MD, coastal ME, MI (south of

43.5° latitude), NJ, NH (south of 43.5° latitude), NY (south of 43.5° latitude), OH, PA,

PR, RI, northern & eastern VA, VI, VT (south of 43.5° latitude), southeastern WI, WV

[22]. Zone II is comprised of Alaska, Hawaii and the rest of the United States which is not located in Zones I or I-A.

3.2 FM Radio Station Classes

The FCC has designated ten different classes of FM radio stations that are identified in Table 3.2-1. Not all of the station classes identified in Table 3.2-1 are authorized for use in all three FM radio zones. Station classes B1 and B are limited to

Zones I and I-A only. Station classes C3, C2, C1, C0 and C are limited to Zone II only

[22]. Station class A is authorized for use in all three zones. Station classes L1 and L2 are reserved for Low Power FM (LPFM) broadcast radio stations. LPFM stations are available to noncommercial, educational, public safety and transportation organizations, but are not available for commercial operations [23]. Station class D is designated for

FM translators that allow stations to provide supplementary service to areas in which direct reception of radio service is unsatisfactory due to distance or intervening terrain barriers [24].

Table 3.2-1: FM Station Classes and Service Contours [1]

Station Maximum Protected Protected Class ERP [kW] HAAT [m] dBu mV/m Contour [km]

A 6.0 100.0 60.0 1.0 28.3

B1 25.0 100.0 57.0 0.7 44.7

B 50.0 150.0 54.0 0.5 65.1

C3 25.0 100.0 60.0 1.0 39.1

C2 50.0 150.0 60.0 1.0 52.2

C1 100.0 299.0 60.0 1.0 72.3

C0 100.0 450.0 60.0 1.0 83.4

C 100.0 600.0 60.0 1.0 91.8

L1 0.1 30.0 60.0 1.0 5.6

L2 0.01 30 60.0 1.0 3.2

D 0.25 N/A 60.0 1.0 N/A

3.3 FM Radio Protected Service Contours

The first column of Table 3.2-1 identifies the station classes. The station classes are distinguished from one another based upon two primary characteristics: the 29

maximum Effective Radiated Power (ERP) and the maximum antenna height. The antenna height is characterized by its effective height above the local area terrain. This is known as Height Above Average Terrain (HAAT). The FCC specifies minimum electric field values to define the protected service area each radio station is allocated.

As specified in Table 3.2-1, the protected service contour is 60 dBu, or 1 mV/m, for all station classes except B1 and B. The protected service contour for B1 and B are 57 dBu and 54 dBu, respectively. The last column of Table 3.2-1 identifies the distance, in kilometers, to the protected service contour. The distance to the protected service contour is a function of ERP and HAAT. In order to predict the distance to the protected service contour, the FCC provides a field strength contour chart known as the F(50,50) field strength contour chart. The F(50,50) field strength contour chart gives the estimated 50% fields strengths exceeded at 50% of the locations in dB above 1 uV/m

[25].

Chapter 4: FM Radio Spectrum Analysis Algorithm

The majority of the verbiage in this chapter, and the subsequent subchapters, are transcribed from [26]. The six-step iterative algorithm that calculates which FM radio stations have protected coverage in a particular area was developed adhering to the rules and regulations governed by the Federal Communications Commission (FCC) Title 47,

Chapter 1, Subchapter C, Part 73, Subpart B – FM Broadcast Stations. The algorithm described in Fig. 3.3-1 is executed on a given coordinate referred to as the Coordinate

Under Analysis (CUA). The algorithm consists of the following six steps:

1. Generate the state station file containing the stations located in the CUA’s state

and all neighboring states.

2. Calculate the distance from each station in the state station file to the CUA.

3. Remove the stations with distances to the CUA outside the maximum protected

service contour.

4. Remove the stations with distances to the CUA outside the protected service

contour for each station class.

5. Calculate each station’s electric field strength at the CUA using a computational

form of the F(50,50) field strength contour chart and each station’s Effective

Radiated Power (ERP) and Height Above Average Terrain (HAAT) [27].

6. Remove the stations with electric field strengths less than the protected electric

field strength for each station class.

Figure 3.3-1: Six-Step Algorithm Flow Diagram [26]. 4.1 Step One: Generate State Station File

The first step of the algorithm generates the state station file from the list obtained from the FCC that contains 37,589 FM radio stations that reside in the United States and the adjacent states in Mexico and Canada. The list was divided into forty-nine individual files representing the forty-eight continental states and Washington, D.C. Each state station file contains the FM radio stations for that respective state and the adjacent states, 32

including those states in Mexico and Canada. For example, the California state station file contains FM radio stations residing in California, Oregon, Nevada, Arizona, Baja

California (Mexico), and Sonora (Mexico).

The list of FM radio stations was divided into forty-nine state files for three reasons. First, the calculated FM radio station data were analyzed on a state-by-state basis. This approach mitigated time consuming re-computation of a single large data file versus smaller data files, in the event that unexpected results were observed. Second, executing the calculations on a state-by-state basis, in lieu of the entire United States at once, significantly reduced the computation processing time. The computation time was significantly reduced because the number of FM radio stations in each of the forty-nine files is significantly smaller than the number of stations in one large file. Third, if any of the forty-nine station state files containing the calculated FM radio station data were to become corrupt, recovering data of a few small files, versus one large file, is less of an impact.

4.2 Step Two: Calculate Station Distances to CUA

The second step of the algorithm iteratively calculates the distance of each FM radio station in the state station file to the CUA using the Haversine equation set (1)

[28].

lon  lon  lon  2 1 lat  lat  lat  2 1 a  sin(lat / 2)2  cos(lat ) cos(lat )sin(lon / 2)2 (1)  1 2  c  2a tan 2( a, 1 a) d  Rc

In (1) the station’s latitude and longitude are represented by lat1 and lon1, respectively and the latitude and longitude of the CUA are represented by lat2 and lon2, respectively. The radius of the Earth, R, is approximated as 6378.1 km.

4.3 Step Three: Remove Stations Outside Maximum Protected Service

Contour

The third step of the algorithm iteratively removes the FM radio stations with a distance to the CUA greater than 91.8 km [29]. No FM radio station, regardless of the station class, is guaranteed protected service outside this maximum protected service contour. As evident in Table 3.2-1, 91.8 km corresponds to the protected service contour of class C stations.

4.4 Step Four: Remove Stations Outside Protected Service Contour

The fourth step of the algorithm iteratively uses the distance values computed in the second step of the algorithm to remove the FM radio stations with distances to the

CUA greater than the values stated in Table 3.2-1 for each radio station class. For example, if a class C3 FM radio station has a distance of 42.1 km to the CUA, the station is removed. This evaluation is performed for the remaining FM radio stations. It is important to note class D FM radio stations are omitted during this step as there is no defined protected service contour for this class. Class D stations are evaluated during the last step of the algorithm.

4.5 Step Five: Calculate FM Stations Field Strength at CUA

The fifth step of the algorithm iteratively calculates each FM radio station’s electric field strength received at the CUA and compares it to the electric field strength values in Table 3.2-1. For a specified distance, the electric field strength of a station is calculated using the F(50,50) field strength contour chart [25]. The F(50,50) contour chart provides field strengths occurring at 50% of the receivers 50% of the time [25].

Using the F(50,50) contour chart to determine the field strengths for this analysis is impractical therefore, the computational form of the F(50,50) provided by [27] is used.

The model approximates the electrical field strength, received power, and median path loss [27].

4.5.1 Propagation Model

The model approximates the electrical field strength, received power, and median path loss [27]. To determine the electric field strength the following calculations are required. The value of the path loss exponent n is calculated using:

4 4 i j n  aij h d (2) i0 j0

Where the coefficients aij are defined in [27] and h, in feet, and d, in miles, represent the antenna height and distance between the FM radio stations and the coordinate under analysis, respectively. Once n has been calculated using (2), the path loss is calculated using:

L  10nlog10 (d)  20log10 (4 / ) dB (3)

Where λ is the wavelength in meters. After the path loss has been calculated using (3), the isotropic received power is calculated using:

Piso  Prad  L dBW (4)

Where Prad is the FM radio station’s ERP in dBW. Lastly, the electric field strength is calculated using:

E   /  480Piso V/m (5)

After the electric field strength has been calculated it is converted to units of dBu for comparison against the electric field strength values, for each station class, in Table

3.2-1.

4.6 Step Six: Remove Stations with Field Strengths Less Than Protected

Field Strengths

The last step of the algorithm iteratively uses the electric field strengths computed in the previous step of the algorithm to remove the stations with electric field strengths less than the strengths stated in Table I for each radio station class. After all field strengths for each station has been evaluated, the stations that have not been removed during any steps in the algorithm have protected coverage at the CUA.

Chapter 5: FM Radio Spectrum Measurements

Before the six-step algorithm that was described in the preceding chapter can be used to determine the unallocated FM radio spectrum across the continental United

States, it must first be validated by comparing it to actual FM radio spectrum measurements. Comparing the results of the six-step algorithm to FM radio spectrum measurements will determine if the algorithm is an accurate representation of the actual

FM radio spectral environment.

5.1 Measurement Locations

To validate the six-step algorithm, measurements of the FM radio station were conducted at the five locations identified in Fig. 5.1-1. The five measurement locations were chosen because they represent various population distributions.

Figure 5.1-1: The five locations where FM radio spectrum measurements were conducted

The first location is in Rockledge, Florida. Rockledge has a population of 24,926 [30].

The second location is the Florida Institute of Technology (FIT) Olin Engineering

Complex, located in the Melbourne-Palm Bay, Florida area which has a population of

76,068 [30]. The third location is in Lake Mary, Florida which is a suburb of Orlando,

Florida. Lake Mary has a population of 13,822 [30]. The fourth location is located in downtown Orlando. Orlando has a population of 238,300 [30]. The fifth location in a rural area outside of St. Cloud, Florida. St. Cloud has a population of 35,183 [30].

5.2 Measurement Hardware

The FM radio spectrum measurements were made using the Diamond D-130NJ antenna and the NooElec RTL-SDR SDR. The D-130NJ antenna operates over the entire UHF and approximately half of the VHF frequency bands. The RTL-SDR

Software Defined Radio (SDR) captured the FM radio spectrum measurements using

MATLAB/Simulink [31]. The RTL-SDR interfaces with the Dell XPS M1330 laptop via the USB.

5.2.1 Receiving Antenna

The Diamond D-130NJ antenna, shown in Fig. 5.2-1, is a discone antenna that exhibits radiation patterns similar to a dipole antenna, has a nominal gain of 2 dBi, and has an operating bandwidth of 50 MHz to 1300 MHz. Since the majority of FM radio receiving antennas are vertically polarized e.g. the receiving antennas used on automobiles, FM radio stations are permitted to transmit using circular or elliptical polarized antennas [32].

FM radio station antennas are primarily circularly polarized to mitigate the excessive polarization loss, on the order of 20 dB, that would occur by having a transmitting antenna that is horizontally polarized and a receiving antenna that is vertically polarized. Because FM radio station broadcasting antennas are typically circularly polarized and the receiving antennas are vertically polarized, the polarization loss is approximately 3 dB. The 3 dB loss occurs because only the vertical component of the circularly polarized electromagnetic wave, which is comprised vertical and horizontal components, is received. Receiving only half of the polarized components results in a loss of 3 dB.

Figure 5.2-1: Diamond D-130NJ antenna The D-130NJ antenna has an N-Type RF connector and the RTL-SDR that will be discussed in the subsequent section, has a MCX connector. In order to interface the

D-130NJ antenna with the RTL-SDR, two cable assemblies were used. The first cable assembly is 10 feet in length and is comprised of LMR-400 RF coaxial cable with two 39

N-Type connectors. LMR-400 has an attenuation of 1.5 dB/100 feet at 150 MHz. The two N-Type connectors have an insertion loss of approximately 0.1 dB each. Therefore, the 10 ft. LMR-400 cable assembly has an insertion loss of 0.35 dB. The second cable assembly, showcased in Fig. 5.2-2, is 18 inches in length and is comprised of PE-

SR405FL RF coaxial cable with N-Type and MCX connectors. This cable assembly has an insertion loss of 0.284 dB at 250 MHz. The overall cable insertion loss is 0.634 dB.

Figure 5.2-2: The RTL-SDR connected to the PE-SR405FL RF coaxial cable assembly 5.2.2 Software Defined Radio

The RTL-SDR that is shown in Fig. 5.2-3 originated from a consumer grade

Digital Video Broadcasting – Terrestrial (DVB-T) USB device that enables users to watch digital television on their computer [31]. The RTL-SDR was not originally designed to be used as a SDR however, engineers and developers in the SDR community unlocked its potential as a programmable SDR [31]. The two primary components of the RTL-SDR are the Rafael Micro R820T digital television tuner and the Realtek

RTL2832U DVB-T Coded Orthogonal Frequency Division Multiplex (COFDM) demodulator [31]. When the COFDM demodulator is put into test mode, the device acts as an Intermediate Frequency (IF) digital radio and outputs raw 8-bit baseband IQ data.

Fig. 5.2-4 is a block diagram of the internal architecture of the R820T and

RTL2832U. The Rafael Micro R820T tuner down converts the RF signal to an IF frequency of 3.57 MHz over a bandwidth of approximately 6 MHz [31]. The signal at the IF frequency is passed to the Realtek RTL2832U for down conversion to baseband

[31].

Figure 5.2-3: Key components of the RTL-SDR [31] The RTL2832U samples the baseband signal at a rate of 28.8 MHz and then performs quadrature demodulation to produce IQ samples [31]. A decimation process is executed to lower the sampling rate to 2.8 MHz and then the samples are routed to a user device over the USB interface [31].

Figure 5.2-4: Internal architecture of the R820T/RTL2832U RTL-SDR [31] The RTL-SDR has a variable gain and a power offset that is a by-product of performing the FFT operation. The variable gain is set in MATLAB/Simulink. The power offset across the FM radio frequency band was determined by inputting a RF signal, generated by a calibrated Keysight Signal Generator, into the RTL-SDR. The input power level and frequency were varied and measured on the RTL-SDR using

MATLAB/Simulink. The averaged FFT power offset measurements yielded 42.34 dB.

5.3 Measurement Software

Measurements of the FM radio spectrum at the five locations were performed using the MathWorks Simulink software suite. To interface the RTL-SDR with

Simulink, the MathWorks RTL-SDR Support from Communications System Toolbox

[33] was used. This toolbox allows the IQ sampled outputs from RTL2832U COFDM demodulator to be easily obtained in Simulink for processing.

Fig. 5.3-1 is the Simulink block diagram that captures the spectral content of a particular frequency within the FM radio band. This block diagram is a modified version of a block diagram provided in [31].

Figure 5.3-1: FM radio spectrum analyzer Simulink software diagram The primary component of the block diagram of Fig. 5.3-1 is the RTL-SDR

Receiver block, which is provided by the RTL-SDR Support from Communications

System Toolbox. The RTL-SDR Receiver block has two inputs and one output. The first input allows the user to specify the RF gain. The RF gain that was used for the measurements presented herein was 20 dB. The second input sets the center frequency that is to be analyzed. In order to quickly increment through each of the 100 FM radio center frequencies, additional logic was used to specify the center frequency. The Index

Vector component allows a center frequency to be chosen based on a FM Channel Select constant. For example a FM Channel Select value of 1 corresponds to a FM radio center frequency of 88.1 MHz and a value of 2 corresponds to 88.3 MHz. The FM Channel

Select’s maximum value is 100, which corresponds to 107.9 MHz. The Index Vector’s center frequency is multiplied by 106, which sets the MHz, and is fed into the RTL-SDR

Receiver Block.

The output of the RTL-SDR Receiver is fed into a high pass filter to remove the unwanted DC frequency and is then fed into a spectrum analyzer component. Fig. 5.3-

2 shows the spectrum analyzer output for FM radio channel 222, which corresponds to a center frequency of 92.3 MHz, at the first location. The spectrum analyzer data is displayed at the baseband frequencies therefore, the FM radio center frequency under analysis corresponds to 0 MHz. The horizontal axis of the spectrum analyzer plot displays the frequency in units of MHz, and has the spectral content of ± 1 MHz around the center frequency. The horizontal axis is in 200 kHz increments, which corresponds to the FM radio station channel size. The vertical axis shows the received power level in units of dBm. The total power of the received signals that are presented in the following section are obtained by integrating the measured power levels from minus

100 kHz to plus 100 kHz around the center frequency. The measured FM radio spectrum is sampled at 2 MHz with a resolution bandwidth of 488 Hz.

Figure 5.3-2: Spectrum analyzer plot of FM radio channel 222 at the first location 5.4 Measurements

To compare the results obtained from the algorithm with the measured results, tables and bar graphs were created that showcase the predicted frequencies alongside the measured frequencies for each of the five locations. The measured Received Signal

Level (RSL) and the predicted RSL for each of the FM radio stations at each location are showcased in the tables and bar graphs. The measured RSL is calculated using (6).

RSLFM  RSLRTLSDR  62.34  0.634  2 dBm (6) 44

In (6), the 62.34 value is the sum of the variable gain, which was set at 20 dB, and the

FFT power offset of 42.34 dB. The 0.634 value is the sum of 0.284 and 0.35. These two values are the insertion losses of the LMR-400 and PE-SR405FL RF cables, respectively. The PE-SR405FL RF cable was used to interface the N-type connector on the LMR-400 RF cable to the MCX connector on the RTL-SDR. The last value, 2, is the nominal gain of the D-130NJ antenna. It is important to note that the algorithm uses a computational form of the F(50,50) contour chart. The F(50,50) predicts field strengths occurring at 50% of the receivers 50% of the time [25]. Because the F(50,50) predicts signal strengths occurring at only 50% of the receivers, there is the other 50% of the receivers where the field strengths could statistically be lower or higher. This is because the F(50,50) contour chart cannot predict variations in terrains, buildings, or any other types of propagation obstacles that could possibly exist between the transmitter and receivers. In addition, the F(50,50) contour chart cannot predict the location of the receiver e.g. inside or outside of a building, which has direct impacts on the received signal levels. Due to these limitations some differences between the predicted and measured received signal strength levels are expected. The purpose of these measurements is not the comparison of the actual received signal levels but the comparison in the actual FM radio station frequencies that were predicted and measured.

The list of 37,589 FM radio stations used in the execution of the algorithm has an associated status field for each station. The status field conveys the current status of each FM radio station e.g. the FM radio station is in the application process (abbreviated

APP), has been granted a construction permit (abbreviated CP), or has been given permission to commence broadcasting (abbreviated Licensed). FM radio stations that

have been designated status APP or CP will show up during the execution of the algorithm but will not be present during the measurements. This is because the station is not actually broadcasting on that respective frequency.

5.4.1 Location 1 Measurements

Figure 5.4-1: Setup of the measurement hardware at location 1 The algorithm and measurement results for the first location are shown in Table

5.4-1 and Fig. 5.4-2. The algorithm for this location predicted 22 channels and the measurements confirmed 17 of the 22 channels. As shown in Table 5.4-1, the 5 frequencies that were not detected, 92.9 MHz, 101.9 MHz, and 105.5 MHz had statuses of CP, APP, and CP, respectively. The two other channel frequencies not detected were

98.5 MHz and 101.1 MHz. This could have been a result of antenna location, terrain or structural blockage, or other propagation factors.

Table 5.4-1: Location 1 FM radio station information

Channel Frequency Predicted RSLdBm Measured RSLdBm Notes

202 88.3 -32.7 -58.9

212 90.3 -45.9 -64.3

214 90.7 -49.7 -79.6

218 91.5 -38.6 -54.2

222 92.3 -31.8 -55.6

225 92.9 -52.4 N/A CP

234 94.7 -26.7 -39.5

236 95.1 -46.9 -66.9

243 96.5 -32.2 -58.5

253 98.5 -47.4 N/A

257 99.3 2.0 -32.3

262 100.3 -31.1 -56.5

266 101.1 -31.0 N/A

270 101.9 -55.7 N/A APP

272 102.3 -28.5 -56.1

274 102.7 -24.8 -54.2

281 104.1 -31.3 -50.0

284 104.7 -50.3 -66.0

286 105.1 -31.3 -48.3

288 105.5 -34.9 N/A CP

292 106.3 -41.5 -50.0

296 107.1 -32.3 -45.6

Figure 5.4-2: Location 1 predicted and measured received signal level strengths The measurements at location 1 detected 6 FM radio stations that the algorithm did not identify. The 6 FM radio stations were 89.3 MHz, 89.5 MHz, 89.8 MHz, 93.9

MHz, 105.9 MHz, and 107.7 MHz.

5.4.2 Location 2 Measurements

Figure 5.4-3: Setup of the measurement hardware at location 2 The measurement results for the second location are shown in Table 5.4-2 and

Fig. 5.4-4. The algorithm for this location predicted 23 channels and the measurements confirmed 22 of the 23 channels. The channel not detected by the measurements had a frequency of 88.3 MHz, and is an educational station that may not have been broadcasting while these measurements were conducted.

Table 5.4-2: Location 2 FM radio station information

Channel Frequency Predicted RSLdBm Measured RSLdBm Notes

202 88.3 -41.0 N/A EDU

208 89.5 9.9 -32.7

212 90.3 -16.4 -68.5

222 92.3 -38.7 -61.9

232 94.3 -39.5 -55.9

236 95.1 -39.0 -46.2

240 95.9 -52.7 -81.5

243 96.5 -39.0 -51.3

250 97.9 -51.4 -74.1

253 98.5 -39.2 -54.1

257 99.3 -42.9 -50.3

259 99.7 -52.9 -69.8

262 100.3 -38.2 -53.8

264 100.7 -50.4 -59.5

266 101.1 -38.1 -54.3

270 101.9 -43.5 -79.5

274 102.7 -51.0 -52.7

279 103.7 -52.6 -66.1

281 104.1 -38.3 -58.4

286 105.1 -38.4 -56.4

292 106.3 -33.9 -45.6

296 107.1 -24.6 -44.6

298 107.5 -52.5 -75.1

Figure 5.4-4: Location 2 predicted and measured received signal level strengths The measurements detected 4 FM radio stations that were not predicted by the algorithm. The 4 FM radio stations were 101.5 MHz, 101.7 MH, 102.3 MHz, and 107.7

MHz. This location is also the same location analyzed in [1], which predicted 24 channels while this paper’s algorithm predicted 23 channels. The channel not predicted or measured by this paper was channel 289, a low power station operating at 105.7 MHz.

It was determined that this station was not predicted because it no longer exists. This was verified by the querying the station’s facility identification number, 193757, in the

FCC’s FM Query Broadcast Station Search tool [34].

5.4.3 Location 3 Measurements

The measurement results for the third location are shown in Table 5.4-3 and Fig.

5.4-5. The algorithm for this location predicted 23 channels and the measurements confirmed 20 of the 23 channels. Two of the three channels that were not detected, frequencies 99.7 MHz and 100.7 MHz, had statuses of APP. The other channel that was not detected was frequency 103.1 MHz.

Table 5.4-3: Location 3 FM radio station information

Channel Frequency Predicted RSLdBm Measured RSLdBm Notes

202 88.3 -38.3 -51.9

205 88.9 -52.9 -79.7

210 89.9 -51.0 -67.3

214 90.7 -39.4 -45.6

220 91.9 -52.5 -44.0

222 92.3 -29.2 -50.8

233 94.5 -24.4 -50.8

237 95.3 -49.2 -69.6

243 96.5 -29.6 -45.6

248 97.5 -54.9 -65.4

251 98.1 -45.3 -51.0

255 98.9 -42.9 -53.7

259 99.7 -43.7 N/A APP

262 100.3 -27.5 -40.6

264 100.7 -52.4 N/A APP

266 101.1 -27.5 -40.7

270 101.9 -15.3 -36.1

276 103.1 -49.2 N/A

281 104.1 -27.7 -44.2

286 105.1 -27.8 -42.5

290 105.9 -15.2 -36.9

294 106.7 -41.8 -55.9

299 107.7 -15.4 -36.5

Figure 5.4-5: Location 3 predicted and measured received signal level strengths The measurements detected 7 FM radio stations that were not predicted by the algorithm. The 7 FM radio stations were 89.3 MHz, 91.5 MHz, 937 MHz, 94.9 MHz,

95.7 MHz, 95.9 MHz, and 96.9 MHz.

5.4.4 Location 4 Measurements

Figure 5.4-6: Setup of the measurement hardware at location 4 53

The measurement results for the fourth location are shown in Table 5.4-4 and

Fig. 5.4-7. The algorithm for this location predicted 26 channels and the measurements confirmed 25 of the 26 channels. The one channel frequency not detected was 98.1

MHz.

Table 5.4-4: Location 4 FM radio station information

Channel Frequency Predicted RSLdBm Measured RSLdBm Notes

202 88.3 -37.0 -38.8

210 89.9 -44.2 -44.3

214 90.7 -37.8 -53.2

218 91.5 -43.9 -50.1

222 92.3 -27.4 -33.3

226 93.1 -34.2 -44.9

233 94.5 -32.9 -54.3

235 94.9 -44.4 -58.6

237 95.3 -29.3 -49.8

240 95.9 -48.0 -59.6

243 96.5 -27.8 -32.9

248 97.5 -45.1 -54.8

251 98.1 -54.6 N/A

255 98.9 -23.4 -44.0

262 100.3 -25.8 -41.6

266 101.1 -25.7 -41.7

270 101.9 -30.4 -44.4

273 102.5 -44.6 -58.9

276 103.1 -39.9 -43.2

281 104.1 -26.0 -49.3

283 104.5 -49.6 -36.7

286 105.1 -26.1 -36.5

290 105.9 -30.4 -51.0

292 106.3 -45.5 -55.2

294 106.7 -32.6 -40.7

299 107.7 -30.5 -54.8

Figure 5.4-7: Location 4 predicted and measured received signal level strengths The measurements detected 3 FM radio stations that were not predicted by the algorithm. The 3 FM radio stations were 89.1 MHz, 89.3 MHz, and 103.7 MHz.

5.4.5 Location 5 Measurements

The measurement results for the fifth location are shown in Table 5.4-5 and Fig.

5.4-8. The algorithm for this location predicted 16 channels and the measurements

confirmed 15 of the 16 channels. The one channel frequency not detected was 90.7

MHz.

Table 5.4-5: Location 5 FM radio station information

Channel Frequency Predicted RSLdBm Measured RSLdBm Notes

202 88.3 -32.7 -62.2

206 89.1 -8.8 -32.7

214 90.7 -49.6 N/A

222 92.3 -32.5 -62.7

243 96.5 -32.9 -57.5

248 97.5 -34.3 -47.3

255 98.9 -55.6 -67.5

262 100.3 -31.3 -48.4

266 101.1 -31.2 -53.6

270 101.9 -45.3 -65.6

281 104.1 -31.4 -54.1

286 105.1 -31.5 -52.0

290 105.9 -45.3 -66.5

294 106.7 -50.9 -67.4

296 107.1 -55.7 -75.7

299 107.7 -45.4 -75.7

Figure 5.4-8: Location 5 predicted and measured received signal level strengths The measurements detected 4 FM radio stations that were not predicted by the algorithm. The 4 FM radio stations were 93.3 MHz, 94.5 MHz, 94.9 MHz, and 98.3

MHz.

5.5 Comparison to Algorithm Results

Examining the bar graphs that were developed for each location, it is evident that the measured received signal power levels were lower than the predicted received signal power levels. This is expected because the F(50,50) contour plots that were used to produce the predicted received signal power levels cannot account for variations in terrain, structural blockage, and other factors that would attenuate a FM radio stations transmitted signal.

On the other end of the spectrum, at each location there were FM radio stations that were measured but were not detected by the algorithm. This is also expected as locations located outside, but near the fringe, of a FM radio station’s service contour receive their low signal strength. This can be understood when tuning stations on a FM radio and experiencing static, yet some intelligible audio broadcasting, on some of the channels.

These unpredicted stations will be accounted for when calculating the vacant FM radio spectrum across the continental United States.

The measurement results of the five locations provided a 94.1% average confidence level that the algorithm predicted the correct FM radio stations that provided service to those specific locations. This provides a high level of confidence that the algorithm is sufficient for estimating the FM radio station coverage across the continental United

States.

Chapter 6: United States FM Radio Maps

The measurements presented in the preceding chapter provide a high level of confidence that the algorithm’s predictions are a good indication of the FM radio stations that have protected coverage at to a particular coordinate. The highly reliable algorithm can now be executed on a large-scale to provide an analysis of the FM radio spectrum across the continental United States. The maps presented herein were developed using Microsoft Map Point.

6.1 Coordinates Under Analysis

The measurements, and the respective algorithm results, presented in the preceding chapter were performed on a single latitude and longitude coordinate. In order to perform a large-scale analysis to predict the FM radio spectrum characteristics across the continental United States, the algorithm must be executed on a large matrix of coordinates. The United States Geological Survey provides a matrix that contains the latitude and longitude component of every human inhabited location in the United States

[35]. Having a matrix of human inhabited locations is preferred over a matrix that contains of mixture of locations that are and are not inhabited by humans. This is because low-power short-range IoT devices and networks are likely to be employed in locations that are inhabited by humans. The United States Geological Survey provides a matrix containing 195,280 human inhabited coordinates across the United States.

6.2 FM Radio Station Protected Coverage Map

The matrix of 195,280 coordinates obtained from [35] was divided into 49 state files, including Washington D.C., to facilitate a state-by-state analysis. The algorithm developed in chapter four of this dissertation was executed on every latitude and longitude coordinate contained in each of the 49 individual state files. The results of the

49 analyses were then plotted across the continental United States using the Microsoft

MapPoint 2013 application. The results of the 49 analyses were averaged per county to provide sufficient granularity. Fig. 6.2-1 shows the predicted FM radio station protected coverage across the continental United States.

Figure 6.2-1: Continental United States FM radio station protected coverage map It is evident from Fig. 6.2-1 that the majority of the counties across the continental

United States have an approximate average of 11 to 25 FM radio stations that have protected coverage in their respective areas. Since there are 100 possible radio channels 60

that could be occupied for every possible latitude and longitude coordinate, on average, the FM radio channels in these counties are 75% to 89% underutilized. Many of the mid-western states such as Kansas, Montana, Nebraska, North Dakota, South Dakota,

Texas, and Wyoming have large areas that have less than 10 FM radio stations. This equates to an underutilization of 90% or more. The counties containing heavily populated urban areas such as New York City, Los Angeles, and Chicago have an average of 35.6, 29.8, 31.8, and 27.5 FM radio stations, respectively. These averages equate to an underutilization of 64.4%, 70.2%, and 68.2%, respectively. It is important to note that the percent underutilization that has been calculated in this section represents the ratio of the number of stations that service an area to the total number of possible FM radio channels, 100. This percent underutilization will differ slightly from the percentage of unallocated FM radio spectrum that is calculated in section 6.3. This is because FM radio stations that broadcast High Definition (HD) radio use 200 kHz more bandwidth than a traditional radio station. This will be discussed in further detail in section 6.3.

In order to identify locations across the continental United States that have underutilized FM radio spectrum that is suitable for secondary licensing of low-power short-range CIoT devices, data of 50 locations with populations ranging from less than

100,000, 100,000 to 1,000,000, and 1,000,000 and more are analyzed in Table 6.2-1 respectively.

Table 6.2-1: Average number of FM radio stations for select locations

Protected SUnallocated SVacant Percentage Location Population Stations [MHz] [MHz] Underutilization

Ruby 360 9.8 17.6 14.6 73.0 Mackinaw 806 19.6 15.7 12.7 63.5 City New Haven 855 9.5 17.1 14.1 70.5 Crosslake 2,141 12.4 17.4 14.4 72.0 Bridgeport 2,409 10.2 17.7 14.7 73.5 Napoleon 8,749 11.1 17 14 70.0 Burley 10,345 7.8 18.4 15.4 77.0 Plainview 22,194 12.1 17.5 14.5 72.5 Richmond 31,364 12.9 16.7 13.7 68.5 Bozeman 37,280 16.8 16.5 13.5 67.5 Richland 48,058 17.6 16.1 13.1 65.5 Chicopee 55,298 15.3 14.6 11.6 58.0 Bismarck 61,727 18 16.2 13.2 66.0 Yuba City 64,925 15.7 14.8 11.8 59.0 Kalamazoo 74,262 20.1 15 12 60.0 Scranton 76,089 21.5 14.3 11.3 56.5 Duluth 86,265 12.1 17.3 14.3 71.5 Lakeland 97,422 20.2 12.8 9.8 49.0 Kenosha 99,218 16.5 14.7 11.7 58.5 Odessa 99,940 18.2 16.4 13.4 67.0 Green Bay 104,057 20.8 15.6 9.6 48.0 Savannah 136,286 26.3 14.3 8.3 41.5 Knoxville 178,874 20.3 14.7 8.7 43.5 Salt Lake 186,440 39.2 9.6 3.6 18.0 City Fort Wayne 253,691 17.7 15.7 9.7 48.5 St. Louis 319,294 27.8 10.4 4.4 22.0 New 343,829 24.3 13 7 35.0 Orleans Wichita 382,368 26.2 12.6 6.6 33.0 Miami 399,457 28.4 10.9 4.9 24.5 Atlanta 420,003 28.5 9.8 3.8 19.0 Las Vegas 583,756 25.2 10 4 20.0 Denver 600,158 26.3 11.3 5.3 26.5 Nashville 601,222 27.9 12.5 6.5 32.5 Washington 601,723 35.7 7.6 1.6 8.0 Seattle 608,660 28.4 10.5 4.5 22.5 Boston 617,594 30.4 9.6 3.6 18.0 Detroit 713,777 36.8 8.1 2.1 10.5 62

Austin 790,390 26.5 12.3 6.3 31.5 San 805,235 33.2 10.1 4.1 20.5 Francisco Jacksonville 821,784 25.5 12.7 6.7 33.5 San Jose 945,942 30 11.1 5.1 25.5 Dallas 1,197,816 30.5 10.1 2.1 10.5 San Diego 1,307,402 25.1 12.9 4.9 24.5 San 1,327,407 24.7 12.4 4.4 22.0 Antonio Phoenix 1,445,632 26.9 11.8 3.8 19.0 Philadelphia 1,526,006 31.9 8.4 0.4 2.0 Houston 2,099,451 27.5 10.8 2.8 14.0 Chicago 2,695,598 31.8 9.2 1.2 6.0 Los 3,792,621 29.8 10 2 10.0 Angeles New York 8,175,133 35.6 9.8 1.8 9.0

The first twenty locations in Table 6.2-1 have populations less than 100,000 and are not a suburb of a major city. All of these locations have less than 22 stations with protected coverage and Scranton, Pennsylvania, has the most stations, 21.5. This equates to a 78.5% underutilization of the channel space by stations with protected coverage. The location in the first publication [1] had a population of 76,068 [30] and had a 76% protected underutilization of the channel space. Based on this analysis smaller cities, towns, and unincorporated locations across the continental United States are expected to have underutilizations of the channel space greater than or equal to 75%.

The next twenty-one locations have populations ranging from 100,000 to

1,000,000. All of these locations have between 20 and 40 stations with protected coverage, with exception to Ft. Wayne Indiana, The location with the most stations is

Salt Lake City, Utah, 39.2 stations. This equates to a 60.8% underutilization of the channel space. In fact, Salt Lake City had the largest average of FM radio stations with

protected coverage in the continental United States. Upon further research it was discovered that Wikipedia states “Salt Lake City has become a case of market saturation on the FM dial; one cannot go through more than about two frequencies on an FM radio tuner before encountering another broadcasting station. A variety of companies, most notably Millcreek Broadcasting and Simmons Media, have constructed broadcast towers on Humpy Peak in the Uinta Mountains to the east. These towers allow frequencies allocated to nearby mountain communities to be boosted by smaller, low- powered FM transmitters along the Wasatch Front” [36]. Based on this description, 1 out of every 3 FM radio channels in Salt Lake City is occupied with a FM radio station.

This equates to an underutilization of approximately two-thirds. This is on par with the

60.8% underutilization that the algorithm predicted.

With the Salt Lake City FM radio dial reportedly being saturated, it is natural to ask the question: if the FM radio dial is saturated, why are the FM channels approximately two-thirds underutilized? This is primarily due to the frequency reuse that is realized by the FCC defining minimum distance separations between FM radio stations. The FCC defines minimum distance separation requirements to mitigate interference between stations. The FCC defines the minimum distance separation between all station classes residing in the same channel and the five adjacent channel pairs [37]. For example, a class A FM radio station is required to be separated from another class A FM radio station at least 115 km, if they are using the same frequency.

If they are using either of the adjacent channel pairs, e.g. +/- 200 kHz, the stations have to be separated at least 72 km. The FCC also defines minimum distance separations for

stations residing in the United States but located near the Mexican border and for stations residing in the United Stations but located near the Canadian border.

The Salt Lake City UT area has the largest number of average FM radio stations within the continental United States. However, it is vital to note that the values in Table

6.2-1 are the average number of FM radio stations that have protected coverage in those respective specific locations. At those locations, there may be additional FM radio stations whose signals may be detected. This is because the F(50,50) field contour plot predicts field strengths occurring at 50% of the receivers 50% of the time [25]. Because the F(50,50) predicts signal strengths occurring at only 50% of the receivers, there is the other 50% of the receivers where the signal strengths could statistically be lower or higher. This is because the F(50,50) contour chart cannot predict variations in terrains, buildings, or any other types of propagation obstacles that could possibly exist between the transmitting stations and receivers. This is why all five measurement locations in section 5.4 had measured stations that weren’t predicated by the algorithm.

Based on the results of this analysis locations across the continental United States with populations ranging from 100,000 to 1,000,000 are expected to have underutilizations less than or equal to 75%. The average number of stations for this population range is 27.8 which equates to a 72.2% underutilization of stations with protected coverage.

The last nine locations in Table 6.2-1 are the locations across the continental

United States with populations greater than 1,000,000 [30]. All of these locations have between 24 and 36 stations with protected coverage. The average number of stations for

this population range is 29.3 which equates to a 70.7% underutilization of stations with protected coverage.

6.3 Unallocated FM Radio Spectrum Coverage Map

The FM radio spectrum ranges from 88 MHz to 108 MHz [20]. The center frequency for each of the FM radio channels starts at 88.1 MHz and increases in increments of 200 kHz up to 107.9 MHz [20]. Therefore, the spectrum range is divided into 100 channels and each channel is allocated 200 kHz of bandwidth for broadcast transmission [20]. FM radio stations that have been approved for In-Band On-Channel

(IBOC) digital radio [38], which provides High Definition (HD) radio broadcasting, utilize 400 kHz of bandwidth. The transmission of the HD digital radio occurs in the sidebands above and below the center frequency [38]. The FM radio station’s digital signal is transmitted in addition to the existing analog signal [38]. Fig. 6.3-1 identifies the two sidebands of a HD FM radio station. The HD radio sidebands consist of 70 kHz of data transmission and 30 kHz for guard bands.

Figure 6.3-1: The spectrum measurement of a HD FM radio station at location 2 The preceding section quantified the average number of FM radio stations that have protected coverage at a respective coordinate. Since select FM radio stations have authorization to transmit digital radio signals along with the traditional analog signals, calculating the allocated FM radio spectrum is more complex than simply multiplying

the number of FM radio stations by 200 kHz. In order to calculate the amount of unallocated FM radio spectrum at a coordinate (7) is used.

SUnallocated  20  (x  y)/5 MHz (7)

Where x is the total number of FM radio stations and y is the number of HD FM radio stations. It is important to note that this complexity builds the case for CIoT. Since the stations may obtain licenses for HD broadcasting whenever they are ready, the situation in any given location may change over time. This necessitates the CIoT equipment that is adaptable to the changing radio conditions. The FCC provides an official list of 1875 radio stations that have been granted authorization to transmit HD FM radio stations alongside traditional analog broadcasting [39]. When the algorithm described in chapter

4 has been executed for a respective coordinate, the resulting data set not only includes the list of stations that are guaranteed coverage in that area, but is also includes the facility ids for each of those respective stations. Each of the facility ids in the resulting data set is compared against the list of 1875 HD FM radio stations. The total number of matching facility ids is equal to y in (7).

To adequately showcase the unallocated FM radio spectrum across the continental

United States, the range of the unallocated FM radio spectrum in Fig. 6.3-2 varies from

0 MHz to 20 MHz. Since the FM radio spectrum ranges from 88 MHz to 108 MHz [20], the total allocated bandwidth is 20 MHz. An area with 0 MHz of unallocated FM radio spectrum would correlate to a completely saturated FM radio spectrum. An area with

20 MHz of unallocated FM radio spectrum would correlate to no FM radio stations servicing that area. The progression from 0 MHz to 20 MHz in Fig. 6.3-2 is represented by the progression from white to green.

Figure 6.3-2: Continental United States unallocated FM radio spectrum map The unallocated FM radio spectrum is inversely related to the total number of

FM radio stations that were showcased in Fig. 6.2-1. More specifically, if an area has a large number of FM radio stations that have protected coverage, the amount of unallocated FM radio spectrum will be small. To maintain a level set analysis, the same locations analyzed in section 6.2 will be analyzed in this section. The first twenty locations of Table 6.2-1 have populations less than 100,000 and are not a suburb of a major city. The unallocated FM radio spectrum in these locations range from 12.8 MHz to 18.4 MHz with an average of 16.2 MHz. Based on the results of this analysis smaller cities, towns, and unincorporated locations across the continental United States are expected to have on average 15 MHz or more of unallocated FM radio spectrum.

The next twenty-one locations of Table 6.2-1 have populations ranging from

100,000 to 1,000,000. The unallocated FM radio spectrum in these locations range from

7.6 MHz to 15.7 MHz with an average of 11.5 MHz. Based on the results of this analysis locations across the continental United States with populations ranging from 100,000 to

1,000,000 are expected to have on average 15 MHz or less of unallocated FM radio spectrum.

The last nine locations of Table 6.2-1 are the locations across the continental

United States with populations greater than 1,000,000 [30]. The unallocated FM radio spectrum in these nine locations range from 8.4 MHz to 12.9 MHz with an average 10.6

MHz.

6.4 Vacant FM Radio Spectrum

The FCC provisions minimum distance separations between stations per station class for the same channel and five pairs of adjacent channels [37]. The purpose of the minimum distance separation is to mitigate interference between stations and to account for the stations with signals that are received outside of their protected service contour.

This was evident when all five locations in chapter 5 measured stations that were not predicted by the algorithm and were therefore, outside of their protected service contour.

The fact that the stations signals propagate outside of their protective service contour provides another argument for CIoT. CR helps avoid interference which may not be predictable or may even exist on a temporary basis as a result of particular propagation conditions. For example, the signal of a given radio station may propagate better in winter when there are no leaves, then in summer when the vegetation is present. In order to obtain the minimum distance separations between stations there will be portions of the spectrum that purposely remain unallocated. The portions of unallocated spectrum will be spread across the 20 MHz of FM radio spectrum. 69

As evident in Table 6.2-1 there exists a correlation between a locations population and the amount of unallocated FM radio spectrum. In addition, the number of stations with signals received outside of their protected service contour will also increase with the locations population. This is expected as the presence of more people increases demand for FM radio stations, particularly HD stations. The analysis indicated that locations with larger populations contain more stations with station class A and L1. The small protected coverage areas of A and L1 allow for the broadcasting of more stations to satisfy the market demand. Many of the largely populated locations in Table 6.2-1 will have much of the unallocated FM radio spectrum occupied by stations with signals received outside of their protected service contour. This can be understood in locations with large populations when tuning an FM radio and having a station on almost every channel. This situation is known as spectrum saturation.

To account for the spectrum used by the stations with signals received outside of their protected service contour, an unpredicted bandwidth α is calculated using (8).

  0.2x  0.2y MHz (8)

Where x' is the total number of unpredicted stations and y' is the subset of the total number of unpredicted stations that are approved for HD broadcasting. Unpredicted stations are defined as stations whose signals are received at locations outside of their protected service contour. The unpredicted bandwidth is proportional to a locations population because x' and y' both increase as the population increases. Chapter 5 conducted measurements at locations with populations ranging from 13,822 to 238,300.

The measurements detected 7 stations that were not predicted by the algorithm. To be conservative, for the populations less than 100,000, x' and y' are set to 10 and 5,

respectively. For this population range, the unpredicted bandwidth is 3 MHz. The next population range, 100,000 to 1,000,000, contains heavily populated locations. Based on the measurements and the results of this analysis, x' and y' are set to 20 and 10, respectively. For this population range, the unpredicted bandwidth is 6 MHz. The last locations are the 9 most populated locations across the continental United States [30].

These locations have populations greater than 1,000,000. These locations have high market demand for FM stations and HD stations, resulting in high volumes of spectrum congestion therefore, x' and y' are equal to 25 and 15, respectively. For this population range, the unpredicted bandwidth is 8 MHz.

The total amount of vacant FM radio spectrum in the three population ranges is calculated using (9).

MHz (9) SVacant  SUnallocated 

The first twenty locations in Table 6.2-1 contain the amount of vacant FM radio spectrum for locations that have populations less than 100,000. The vacant FM radio spectrum in these locations range from 9.8 MHz to 15.4 MHz with an average of 13.2

MHz. Based on the results of this analysis smaller cities, towns, and unincorporated locations across the continental United States are expected to have on average 13 MHz or more of vacant FM radio spectrum.

The next twenty-one in Table 6.2-1 contains FM have populations ranging from

100,000 to 1,000,000. The vacant FM radio spectrum in these locations range from 3.6

MHz to 11.7 MHz with an average of 5.5 MHz. Based on the results of this analysis locations across the continental United States with populations ranging from 100,000 to

1,000,000 are expected to have on average 6 MHz or less of vacant FM radio spectrum.

The last nine locations in Table 6.2-1 are the locations across the continental

United States with populations greater than 1,000,000 [30]. The vacant FM radio spectrum in these nine locations range from 0.4 MHz to 4.9 MHz with an average 2.7

MHz.

6.5 CIoT Deployment Bitrates

Predicting the bitrates of a CIoT deployment is challenging because each deployment has unique design parameters such as receiver sensitivity, required bit error rate, etc. and an RF environment that varies greatly with each location. Therefore, the bitrates discussed in this section assume that the design engineer has adequately assessed the design parameters and RF environments to achieve the SNRs required to realize these bitrates with the respective modulation scheme.

Table 6.5-1: Potential CIoT deployment bitrates

Bitrate [Mbps] β=1 β=0.75 β=0.50 β=0.25 SVacant Location Population 8- 8- 8- 8- [MHz] QPS PS QPSK PS QPSK PS QPSK PS K K K K K Ruby 360 14.6 29.2 43.8 21.9 32.9 14.6 21.9 7.3 11.0 Mackinaw 806 12.7 25.4 38.1 19.1 28.6 12.7 19.1 6.4 9.5 City New Haven 855 14.1 28.2 42.3 21.2 31.7 14.1 21.2 7.1 10.6 Crosslake 2,141 14.4 28.8 43.2 21.6 32.4 14.4 21.6 7.2 10.8 Bridgeport 2,409 14.7 29.4 44.1 22.1 33.1 14.7 22.1 7.4 11.0 Napoleon 8,749 14 28.0 42.0 21.0 31.5 14.0 21.0 7.0 10.5 Burley 10,345 15.4 30.8 46.2 23.1 34.7 15.4 23.1 7.7 11.6 Plainview 22,194 14.5 29.0 43.5 21.8 32.6 14.5 21.8 7.3 10.9 Richmond 31,364 13.7 27.4 41.1 20.6 30.8 13.7 20.6 6.9 10.3 Bozeman 37,280 13.5 27.0 40.5 20.3 30.4 13.5 20.3 6.8 10.1 Richland 48,058 13.1 26.2 39.3 19.7 29.5 13.1 19.7 6.6 9.8 Chicopee 55,298 11.6 23.2 34.8 17.4 26.1 11.6 17.4 5.8 8.7 Bismarck 61,727 13.2 26.4 39.6 19.8 29.7 13.2 19.8 6.6 9.9

Yuba City 64,925 11.8 23.6 35.4 17.7 26.6 11.8 17.7 5.9 8.9 Kalamazoo 74,262 12 24.0 36.0 18.0 27.0 12.0 18.0 6.0 9.0 Scranton 76,089 11.3 22.6 33.9 17.0 25.4 11.3 17.0 5.7 8.5 Duluth 86,265 14.3 28.6 42.9 21.5 32.2 14.3 21.5 7.2 10.7 Lakeland 97,422 9.8 19.6 29.4 14.7 22.1 9.8 14.7 4.9 7.4 Kenosha 99,218 11.7 23.4 35.1 17.6 26.3 11.7 17.6 5.9 8.8 Odessa 99,940 13.4 26.8 40.2 20.1 30.2 13.4 20.1 6.7 10.1 Green Bay 104,057 9.6 19.2 28.8 14.4 21.6 9.6 14.4 4.8 7.2 Savannah 136,286 8.3 16.6 24.9 12.5 18.7 8.3 12.5 4.2 6.2 Knoxville 178,874 8.7 17.4 26.1 13.1 19.6 8.7 13.1 4.4 6.5 Salt Lake 186,440 3.6 7.2 10.8 5.4 8.1 3.6 5.4 1.8 2.7 City Fort Wayne 253,691 9.7 19.4 29.1 14.6 21.8 9.7 14.6 4.9 7.3 St. Louis 319,294 4.4 8.8 13.2 6.6 9.9 4.4 6.6 2.2 3.3 New 343,829 7 14.0 21.0 10.5 15.8 7.0 10.5 3.5 5.3 Orleans Wichita 382,368 6.6 13.2 19.8 9.9 14.9 6.6 9.9 3.3 5.0 Miami 399,457 4.9 9.8 14.7 7.4 11.0 4.9 7.4 2.5 3.7 Atlanta 420,003 3.8 7.6 11.4 5.7 8.6 3.8 5.7 1.9 2.9 Las Vegas 583,756 4 8.0 12.0 6.0 9.0 4.0 6.0 2.0 3.0 Denver 600,158 5.3 10.6 15.9 8.0 11.9 5.3 8.0 2.7 4.0 Nashville 601,222 6.5 13.0 19.5 9.8 14.6 6.5 9.8 3.3 4.9 Washington 601,723 1.6 3.2 4.8 2.4 3.6 1.6 2.4 0.8 1.2 Seattle 608,660 4.5 9.0 13.5 6.8 10.1 4.5 6.8 2.3 3.4 Boston 617,594 3.6 7.2 10.8 5.4 8.1 3.6 5.4 1.8 2.7 Detroit 713,777 2.1 4.2 6.3 3.2 4.7 2.1 3.2 1.1 1.6 Austin 790,390 6.3 12.6 18.9 9.5 14.2 6.3 9.5 3.2 4.7 San 805,235 4.1 8.2 12.3 6.2 9.2 4.1 6.2 2.1 3.1 Francisco Jacksonville 821,784 6.7 13.4 20.1 10.1 15.1 6.7 10.1 3.4 5.0 San Jose 945,942 5.1 10.2 15.3 7.7 11.5 5.1 7.7 2.6 3.8 Dallas 1,197,816 2.1 4.2 6.3 3.2 4.7 2.1 3.2 1.1 1.6 San Diego 1,307,402 4.9 9.8 14.7 7.4 11.0 4.9 7.4 2.5 3.7 San Antonio 1,327,407 4.4 8.8 13.2 6.6 9.9 4.4 6.6 2.2 3.3 Phoenix 1,445,632 3.8 7.6 11.4 5.7 8.6 3.8 5.7 1.9 2.9 Philadelphia 1,526,006 0.4 0.8 1.2 0.6 0.9 0.4 0.6 0.2 0.3 Houston 2,099,451 2.8 5.6 8.4 4.2 6.3 2.8 4.2 1.4 2.1 Chicago 2,695,598 1.2 2.4 3.6 1.8 2.7 1.2 1.8 0.6 0.9 Los Angeles 3,792,621 2 4.0 6.0 3.0 4.5 2.0 3.0 1.0 1.5 New York 8,175,133 1.8 3.6 5.4 2.7 4.1 1.8 2.7 0.9 1.4

Table 6.5-1 contains the bitrates for the locations in Table 6.2-1 using modulation types, Quadrature Phase Shift Keying (QPSK) and 8-PSK. These two modulation types use the vacant FM radio spectrum multiplied by the bandwidth utilization factor β to generate the potential bitrates for low-power short-range CIoT deployments. The bandwidth utilization factor ranges between 0 and 1 and represents the percentage of bandwidth being used for data transmission. The bandwidth not being used for data transmission is used for guard bands, filtering, etc. For the purposes of this analysis the bandwidth utilizations of β = 0.75 and β=0.5 are used.

The locations in Table 6.5-1 with populations less than 100,000 yield average bitrates of 19.8 Mbps, 29.7 Mbps, 13.2 Mbps, and 19.8 Mbps for QPSK and 8-PSK with

β = 0.75 and QPSK and 8-PSK with β = 0.5, respectively.

The locations in Table 6.5-1 with populations between 100,000 and 1,000,000 yield average bitrates of 8.3 Mbps, 12.5 Mbps, 5.5 Mbps, and 8.3 Mbps for QPSK and

8-PSK with β = 0.75 and QPSK and 8-PSK with β = 0.5, respectively.

The nine locations in Table 6.5-1 that have populations greater than 1,000,000 yield average bitrates of 3.9 Mbps, 5.9 Mbps, 2.6 Mbps, and 3.9 Mbps for QPSK and

8-PSK with β = 0.75 and QPSK and 8-PSK with β = 0.5, respectively.

6.6 FM Radio Map Conclusions

Sections 6.2 through 6.5 quantify the average number of FM radio stations that have protected coverage, the unallocated FM radio spectrum, the vacant FM radio spectrum, and the potential CIoT bitrates across the continental United States, respectively.

Section 6.2 not only quantifies the average number of FM radio stations across the continental United States, reference Fig. 6.2-1, but it also provides a detailed analysis 74

of select locations, of varying population densities, to show the relationship between population and the number of FM radio stations. Table 6.2-1 contains the population and average number of FM radio stations for 50 select locations. The 50 locations were randomly chosen to represent areas that have populations ranging from 360 to

8,175,133. The population of 8,175,133 is the population of New York City and represents the largest population of any area in the United States. The area in the continental United States that has the largest average number of FM radio stations is

Salt Lake City Utah. This area has an average of 39.2 FM radio stations. This equates to an underutilization of 61%.

Section 6.3 builds upon the analysis of section 6.2 by quantifying the amount of unallocated FM radio spectrum across the continental United States. Fig. 6.3-2 provides a map that quantifies the amount of unallocated FM radio spectrum across the United

States. It has been shown in Fig. 6.3-2 that in the more populated cities there exists unallocated FM radio spectrum on the order of 10 MHz. In rural less populated areas, the unallocated spectrum is on the order of 15 MHz to 18 MHz. Table 6.3-1 verifies this by quantifying the actual amount of unallocated FM radio spectrum for the 50 select locations in Table 6.2-1.

Section 6.4 uses the results of section 6.3 to calculate the vacant FM radio spectrum at the locations in Table 6.2-1. The locations with populations less than

100,000 have vacant FM radio spectrum ranging from 9.8 MHz to 15.4 MHz with an average of 13.2 MHz. The locations with populations ranging from 100,000 to

1,000,000 have vacant FM radio spectrum ranging from 3.6 MHz to 11.7 MHz with an average of 5.5 MHz. The last nine locations are the locations across the continental

United States with populations greater than 1,000,000 [30] and have vacant FM radio spectrum ranging from 0.4 MHz to 4.9 MHz with an average 2.7 MHz.

Section 6.5 finalizes the FM radio spectrum analysis by quantifying the potential

CIoT bitrates of the locations in Table 6.5-1. The locations with populations less than

100,000 yielded average bitrates of 19.8 Mbps, 29.7 Mbps, 13.2 Mbps, and 19.8 Mbps for QPSK and 8-PSK with β = 0.75 and QPSK and 8-PSK with β = 0.5, respectively.

The locations with populations between 100,000 and 1,000,000 yielded average bitrates of 8.3 Mbps, 12.5 Mbps, 5.5 Mbps, and 8.3 Mbps for QPSK and 8-PSK with β = 0.75 and QPSK and 8-PSK with β = 0.5, respectively. The last nine locations have populations greater than 1,000,000 yielded average bitrates of 3.9 Mbps, 5.9 Mbps, 2.6

Mbps, and 3.9 Mbps for QPSK and 8-PSK with β = 0.75 and QPSK and 8-PSK with β

= 0.5, respectively.

Chapter 7: CIoT Deployment in the FM Radio Spectrum

Chapter 6 quantified the average number of FM radio stations that have protected coverage, the unallocated FM radio spectrum, the vacant FM radio spectrum, and the potential CIoT bitrates across the continental United States. With the potential CIoT bitrates across the continental United States defined, the deployment of CIoT devices in the FM radio spectrum can now be explored. As briefly discussed in chapter 6, the potential CIoT bitrates are aggregated rates. Aggregated in this sense means that the bitrates have not been divided or limited to facilitate more than one CIoT device pair in the same portion of FM radio spectrum. In practical IoT applications, multiple CIoT devices will be used therefore, each deployment will need to balance the number of

CIoT devices with the required spectrum bandwidth allocations to meet the specific bitrate requirements.

To assess the deployment of CIoT devices in the vacant FM radio spectrum three topics will be discussed. The first topic to be discussed is the characteristics of low- power short-range CIoT devices that will occupy the vacant FM radio spectrum. The second topic to be discussed is how to utilize the vacant FM radio spectrum. The third topic to be discussed is how to control access to the vacant spectrum. The second and third topics provide a concrete spectrum management approach.

7.1 Low-Power Short-Range CIoT Devices

Utilization of the vacant FM radio spectrum for secondary licensing via CR is not limited to low-power short-range devices. However, due to the characteristics of IoT applications, devices operating in low power modes at short ranges achieve the most

benefit. For low-power short-range IoT devices to use the vacant FM radio spectrum,

CR technology must be implemented, hence CIoT devices. The majority of CIoT devices are battery powered and do not have the ability to draw from shore power [9].

In addition, CIoT devices are often deployed in locations that are not easily accessible for battery replacement. Therefore, CIoT devices must operate in low-power modes over long periods of time. As discussed in the literature review cognitive M2M, or more generally CIoT devices, are divided into two classifications capillary and cellular [9].

Both capillary and cellular devices require the use low-power technologies [9]. In addition, since most capillary devices use technologies such as ZigBee and Wi-Fi, they are also limited to short-range applications [9].

7.2 Spectrum Management: Access Control

It is important to manage and control access to the vacant FM radio spectrum across the continental United States to protect and mitigate interference to the primary users, the FM radio stations. If interference to the FM radio stations is not mitigated, the

FCC will not allow secondary use of the vacant FM radio spectrum. The FCC has already recognized the potential that the TVWS has for secondary licensing. The TVWS regulations have laid the foundation for regulations that can be applicable to future frequency bands of interest, such as the FM radio spectrum. The FCC has defined three regulations for protecting the primary users of the TV channels from secondary users

[15]:

a. Utilization of a TVWS database that contains the vacant TV channels for a

specific area

b. Adherence to transmit power limitations 78

c. Ability to perform real-time TV channel spectrum sensing

To provide sound spectrum access control for the vacant FM radio spectrum, these same

TVWS regulations structure can be implemented. Though the TVWS regulations provide interference mitigations for the primary user, they do not provide regulations on protecting secondary TVWS device user from interfering with each other. It is important that both the primary and secondary users are protected from interference.

7.2.1 FM Radio Spectrum Geolocation Database

The FM radio spectrum data generated for the continental United States in chapter 6 can be used to populate a database that contains the number of FM radio stations and the corresponding vacant FM radio spectrum for a respective area. The FM radio spectrum data for specific geo-graphical areas can be queried based on latitude and longitude coordinates. Therefore, CIoT devices will need to have an embedded GPS device to pinpoint their location and determine which FM radio channels are vacant and can be used for transmission.

Two methods can be used to access the vacant FM radio spectrum database. The first method is having a local copy of the vacant FM radio spectrum database located within the CIoT device. The local copy of the vacant FM radio spectrum database would be periodically updated. This method is advantageous because the FM radio stations that service a specific geo-graphical location do not regularly change. This method also mitigates the need to have consistent access to the remote maser vacant FM radio spectrum database. This method is ideal for CIoT devices that are used in mobile, non- stationary, applications. The second method is geared towards stationary CIoT devices that would have regular access to the remote master vacant FM radio spectrum database. 79

Since the FM radio spectrum footprint does not regularly change, the CIoT device using this method could access the remote master vacant FM radio spectrum database upon device power on.

Ultimately, the method of access to the vacant FM radio spectrum database will be driven by the application, mobile or non-stationary, of the CIoT device. Regardless of the access method, each CIoT device will need a back-up ability to sense the FM radio spectrum to determine which FM radio channels can be used for transmission in the event that vacant FM radio spectrum database cannot be accessed.

7.2.2 Spectrum Sensing

The preferred method of determining which FM radio channels are vacant and can be used for transmission would be accessing the vacant FM radio spectrum database.

However, to ensure that the primary users are protected from interference, the back-up method of determining which FM radio channels are vacant must be implemented. The back-up method would be employed in the event that access to the vacant FM radio spectrum database could not be established. Since the CIoT devices will use CR technology, the back-up method would be the ability of the CIoT device to sense the spectrum.

Spectrum sensing is a primary component of CR technology. Spectrum sensing is the ability of a CR device to detect, in real-time, the presence of primary users in a given frequency band. Spectrum sensing is the subject of ongoing research in both academia and industry and multiple approaches have been developed. One spectrum sensing algorithm that could be implemented in CIoT devices is the energy detection algorithm. 80

The energy detection algorithm involves the discrimination between samples that contain only noise and samples that contain signal power with high noise power

[40]. In energy detection, the energy of the received signal samples is compared to a fixed threshold to determine the presence or absence of primary users [40]. When the energy detection algorithm is executed on the FM radio spectrum, the threshold value should theoretically depend on the radio class of the station that may or may not be occupying a particular radio channel. Since the FM radio class cannot be known a priori, as this would defeat the purpose of detecting the presence of a radio station, the largest protected electrical field strength must be assumed. The maximum protected electric field strength for all of the FM radio station classes is 60 dBuV/m or 60 dBu [22]. The energy detection algorithm will be comparing sampled signal power levels therefore, the 60 dBu electric field strength value needs to be converted to an equivalent RSL per the following equation for a 0 dBi receive antenna:

RSL  20log E  20log f  77.2 dBm (8) channel max,uV / m channel,MHz

Equation (8) computes the RSL value that corresponds to the maximum protected field strength value for a specific channel. It is important to note that the conversion from the electrical field strength to RSL is a function of frequency.

Therefore, each channel will have a unique RSL as the energy detection algorithm’s threshold value. The energy detection algorithm is based on the evaluation of two hypotheses [40]:

(9) H0 : Y[n] = W[n]

(10) H1:Y[n] = X[n]+ W[n]

The first hypothesis (9) is the absence of a primary users signal and only the noise signal 푊[푛] is present [40]. The second hypothesis (10) is the existence of the primary users signal 푋[푛] in addition to the noise signal. The decision as to which hypothesis is valid is based on the comparison of the decision statistic to the threshold value [40]. The decision statistic is calculated using the following equation [40]:

1 N 2 T   Y[n] (11) N i1

The decision statistic (11) is equal to the average of all of the magnitude squared samples [40]. Upon calculation of the decision statistic, it is compared to the threshold value using the following relationships:

(12) H0  True if T < RSLchannel

(13) H1  True if T  RSLchannel

If the decision statistic is less than the RSL threshold value for the respective

FM radio channel then there only exists the noise signal and the radio channel can be considered vacant. Once the FM radio channel has been deemed vacant, it can be used for transmission. If the decision statistic is greater than or equal to the RSL threshold value then the FM radio channel will be considered occupied and it cannot be used for transmission.

7.3 Spectrum Management: Spectrum Utilization

7.3.1 Frequency Multiplexing

The vacant FM radio spectrum in contains unused channels, with corresponding unused center frequencies that range from 88 MHz to 108 MHz [20], which are non-

contiguous in nature. For a CIoT deployment to capitalize on the most vacant FM radio spectrum, a frequency multiplexing scheme that can transmit on non-contiguous frequency carriers is required. Using non-contiguous vacant FM radio spectrum can be accomplished through a multiplexing scheme such as the Non-Contiguous Orthogonal

Frequency Division Multiplexing (NC-OFDM) scheme. The NC-OFDM scheme allows data to be multiplexed on non-continuous sub-carriers. This will allow data to be transmitted on the non-continuous 200 kHz vacant FM radio channels. For example,

Fig. 7.3-1 contains a measured spectrum plot of eight FM radio channels centered on channel 222 at location 1. Channels 218 and 222 contain an analog FM radio station and a HD radio station, respectively. Therefore, channels 219, 220, 224, 225 and 226 are non-contiguous vacant channels that can be used for bi-directional transmissions between CIoT devices.

Figure 7.3-1: Non-contiguous vacant FM radio channels at location 1 7.3.2 Power Management

To help ensure interference to the FM radio stations and other secondary CIoT devices is mitigated, utilization of a vacant FM radio spectrum database and an energy detection spectrum sensing system, for back-up situations, can be used. The vacant FM radio spectrum database, and back-up spectrum sensing system, provide the transmission channels by making the decision whether the FM radio channel, or another secondary CIoT device, is occupying those specific channels. Once the decision of 83

channel occupancy has been made, a proper power management is employed to ensure the following:

a. FM radio stations in adjacent radio channels are not subjected to interference.

Note that sufficient frequency guard bands need to be established and filtering

needs to be implemented in the transmission path to aid in limiting interference

to adjacent radio channels.

b. Inference between secondary CIoT devices is mitigated.

c. Reuse of the vacant FM radio channels by multiple secondary CIoT devices can

be achieved and will increase the number of CIoT devices able to operate.

Though it has been previously stated, it is of utmost importance to mitigate interference to the primary user, the FM radio stations. Therefore, once a FM radio channel has been determined to be vacant and can be used for transmission that does not mean that the secondary CIoT device is free to transmit at any power level. The

CIoT device would be required to limit its transmit power level to mitigate inference to adjacent FM radio channels. The mitigation of interference is not only achieved by limiting transmission power but also by implementing guard bands and implementing a transmission filter. The transmission power will also be limited to facilitate reuse of the vacant FM radio channels between multiple CIoT devices. By reusing vacant FM radio channels, the effective capacity of each vacant radio channel increases at a rate directly proportional to the number of CIoT devices using that radio channel.

Chapter 8: Summary and Conclusions

8.1 Research Summary

The rapid deployment of the IoT has introduced a new class of low-power short- range wireless devices are placing an overwhelming demand on the radio spectrum. This overwhelming demand is creating a shortage of radio spectrum required for new wireless devices. To combat the shortage of radio spectrum, IoT devices need to be able to identify vacant radio spectrum that can be used in an opportunistic manner through

CR. Chapter 1 established the following research question and the impact of the research: How can vacant FM radio spectrum be used through CR to optimize the throughput of low-power short-range IoT systems while minimizing interference to the

FM radio stations (primary users)?

Chapter 2 provided a thorough review of the literature to formulate a sound understanding of the research question. The research question can be divided into two primary areas of focus, IoT and CR. Extensive review of the literature concluded that other than this dissertation’s published research [1] and [26], there are no publications pertaining to the use of the FM radio spectrum, via CR, for IoT applications.

Furthermore, there is no published work that analyzes the FM radio spectrum in order to fully understand its vast potential and capacity for CR applications.

Chapter 3 provided an overview of the fundamental details of FM radio that is required for the analysis of the FM radio spectrum.

Chapter 4 described the six-step iterative algorithm that calculated the coverage of the FM radio stations across the continental United States. The algorithm was developed

while adhering to the rules and regulations governed by the FCC. The iterative algorithm is comprised of six steps that are executed for a given latitude and longitude coordinate.

The first step of the algorithm develops the state station files used in the analysis. The second step of the algorithm calculates the distances from each radio station to the CUA.

The third step of the algorithm removes the radio stations outside of the maximum protected service contour. The fourth step of the algorithm removes the radio stations outside of the protected service contour per radio station class. The fifth step of the algorithm calculated the each FM radio station’s electric field strength at the CUA. The sixth and last step of the algorithm removes stations with field strengths less than the protected field strengths per station class.

Prior to quantifying the vacant FM radio spectrum across the continental United

States, measurements were conducted in chapter 5 to validate the algorithm presented in chapter 4. To validate the algorithm, measurements at five strategically chosen locations were conducted. The five measurement locations were chosen to represent various population distributions. The first location was located in Rockledge, Florida and has a population of 24,926 [30]. The second location is the FIT Olin Engineering

Complex, located in the Melbourne-Palm Bay, Florida area which has a population of

76,068 [30]. The third location is in Lake Mary, Florida which is a suburb of Orlando,

Florida and has a population of 13,822 [30]. The fourth location is located in downtown

Orlando and has a population of 238,300 [30]. The fifth location is a rural area outside of St. Cloud, Florida and has a population of 35,183 [30]. The measurement results of the five locations provided a 94.1% average confidence level that the algorithm predicted the correct FM radio stations that provided service to those specific locations.

This provided a high level of confidence that the algorithm was sufficient for estimating the FM radio station coverage across the continental United States.

Chapter 6 quantified the average number of FM radio stations that have protected coverage, the unallocated FM radio spectrum, the vacant FM radio spectrum, and the potential CIoT bitrates across the continental United States, respectively. Chapter 6 not only quantifies the FM radio stations characteristics across the continental United States but it also provides a detailed analysis of select areas, of varying population densities, to show the relationship between population and the number of FM radio stations. Fifty select areas were randomly chosen to represent areas that have populations ranging from

360 to 8,175,133. The population of 8,175,133 is the population of New York City and represents the largest population of any area in the United States. The area in the continental United States that has the largest average number of FM radio stations is

Salt Lake City Utah. This area has an average of 39.2 FM radio stations. This equates to an underutilization of 61%.

Chapter 7 discussed the deployment of CIoT devices in the vacant FM radio spectrum and provides a concrete spectrum management approach. The first of three topics discussed was the characteristics of low-power short-range CIoT devices that will occupy the vacant FM radio spectrum. The second topic that was discussed is how to utilize the vacant FM radio spectrum. The third topic that was discussed is how to control access to the vacant spectrum.

8.2 Conclusions

The research in this dissertation has shown that of the three population groups, locations with a population less than 100,000, and were not located near a major heavily 87

populated location, showed the most potential for secondary licensing of CIoT devices.

Locations with a population less than 100,000 yielded on average at least 13 MHz of vacant FM radio spectrum. The analysis also indicated that locations with populations greater than 1,000,000 did not yield enough vacant spectrum to justify risking interference to other FM radio stations.

To achieve the full potential of the vacant spectrum, CIoT devices must operate in a low-power short-range mode to facilitate frequency reuse amongst CIoT devices.

Low-power short-range CIoT devices using frequency-reuse can obtain potential bitrates of approximately 30 Mbps. A spectrum management plan that address controlling access and utilization of the vacant spectrum was discussed. To manage access to the vacant spectrum two methods were presented, a FM Radio Spectrum

Geolocation Database, similar to the database that manages access to the TVWS, and spectrum sensing technology, which is a fundamental concept of CR. To manage utilization of the vacant spectrum frequency multiplexing and proper power management schemes were discussed.

The results of the research presented in this dissertation answer the principle research question: How can vacant FM radio spectrum be used through CR to optimize the throughput of low-power short-range IoT systems while minimizing interference to the FM radio stations (primary users)?

References

[1] D. T. Otermat, C. E. Otero and I. Kostanic, "Analysis of the FM Radio Spectrum for Internet of Things Opportunistic Access via Cognitive Radio," in 2015 IEEE 2nd World Forum on Internet of Things (WF-IoT), Milan, 2015.

[2] J. A. Stankovic, "Research Directions for the Internet of Things," IEEE Internet of Things Journal, vol. 1, no. 1, pp. 3-9, 2014.

[3] W. Li, C. Zhu, V. C. M. Leung, L. T. Yang and Y. Ma, "Performance Comparison of Cognitive Radio Sensor Networks for Industrial IoT With Different Deployment Patterns," IEEE Systems Journal, vol. PP, no. 99, pp. 1-11, 2015.

[4] L. Tan and N. Wang, "Future Internet: The Internet of Things," in 3rd International Conference on Advanced Computer Theory and Engineering (ICACTE), Chengdu, 2010.

[5] D. Singh, G. Tripathi and A. J. Jara, "A survey of Internet-of-Things: Future Vision, Architecture, Challenges and Services," in IEEE World Forum on Internet of Things (WF-IoT), Seoul, 2014.

[6] L. Coetzee and J. Eksteen, "The Internet of Things - Promise for the Future? An Introduction," in IST-Africa 2011 Conference Proceedings, 2011.

[7] A. Iera, C. Floerkemeier, J. Mitsugi and G. Morabito, "The Internet of Things," IEEE Wireless Communications, pp. 8-9, 2010.

[8] M. H. Asghar, A. Negi and N. Mohammadzadeh, "Principle Application and Vision in Internet of Things (IoT)," in International Conference on Computing, Communication and Automation (ICCCA2015), Uttar Pradesh, 2015.

[9] A. Aijaz and A. A. H, "Cognitive Machine-to-Machine Communications for Internet-of-Things: A Protocol Stack Perspective," IEEE Internet of Things Journal, vol. 2, no. 2, pp. 103-112, 2015.

[10] Y. Zhang, R. Yu, M. Nekovee, L. Y and S. G. S. Xie, "Cognitive Machine-to- Machine Communications: Visions and Potentials for the Smart Grid," IEEE Network Journal, pp. 6-13, 2012.

[11] President's Council of Advisors on Science and Technology, "Realizing the Full Potential of Government-Held Spectrum to Spur Economic Growth," Washington D.C., 2012.

[12] D. A. Roberson, C. S. Hood, J. L. LoCicero and J. T. MacDonald, "Spectral Occupandy and Interference Studies in suppor of Cognitive Radio Technology Deployment," in IEEE Workshop on Network Technologies for Software Defined Radio Networks, Reston, 2006.

[13] S. W. Ellingson, "Spectral Occupancy at VHF: Implications for Freqeuency-Agile Cognitive Radio," in IEEE Vehicular Technology Conference, Dallas, 2005.

[14] S. Haykin, "Cognitive Radio: Brain-Empowered Wireless Communications," IEEE Journal on Selected Areas in Communications, vol. 20, no. 2, pp. 201-220, 2005.

[15] T. Baykas, M. Kasslin, M. Cummings, H. Kang, J. Kwak, R. Paine, A. Reznik, R. Saeed and S. J. Shellhammer, "Developing a Standard for TV White Space Coexistence: Technical Challenges and Solution Approaches," IEEE Wireless Communications, 2012.

[16] D. Makris, G. G. Gardikis and A. Kourtis, "Quantifying TV White Space Capacity: A Geolocation-Based Approach," IEEE Communications Magazine, September 2012.

[17] K. Harrison, S. M. Mishra and A. Sahai, "How much white-space capacity is there?," in IEEE DySpan, Singapore, 2010.

[18] P. Palka and P. Neumann, "Analyzing the Availability of TV White Spaces in Dynamic Broadcast," IEEE Transactions on Consumer Electronics, vol. 60, no. 3, pp. 302-210, 2014.

[19] G. P. Villardi, Y. D. Alemseged, C. Sun, C. Sum, T. Nguyen, T. Baykas and H. Harada, "Enabling Coexistance of Multiple Cognitive Networks in TV White Space," IEEE Wireless Communications, pp. 32-40, 2011.

[20] Federal Communications Commission, 47CFR1.C.73.201 "Numerical designation of FM broadcast channels".

[21] Federal Communications Commission, 47CFR1.C.73.202 "Table of Allotments".

[22] Federal Communications Commission, "FM Broadcast Station Classes and Service Contours," 5 November 2014. [Online]. Available: http://www.fcc.gov/encyclopedia/fm-broadcast-station-classes-and-service- contours.

[23] Federal Communications Commission, "Low Power FM Broadcast Radio Stations (LPFM)," 5 November 2014. [Online]. Available: http://www.fcc.gov/encyclopedia/low-power-fm-broadcast-radio-stations-lpfm.

[24] Federal Communications Commission, "FM Translators and Boosters - General Information," [Online]. Available: http://www.fcc.gov/encyclopedia/fm- translators-and-boosters-general-information.

[25] Federal Communications Commission, 47CFR1.C.73.313 "Prediction of Coverage".

[26] D. T. Otermat, I. Kostanic and C. E. Otero, "Analysis of the FM Radio Spectrum for Secondary Licensing of Low-Power Short-Range Cognitive Internet of Things Devices," IEEE Access, vol. PP, no. 99, pp. 1 - 1, 2016.

[27] C. Perez-Vega and J. M. Zamanillo, "Path-Loss Model for Broadcasting Applications and Outdoor Communication Systems in the VHF and UHF Bands," IEEE Transactions on Broadcasting, vol. 48, no. 2, pp. 91-96, 2002.

[28] A. G. Ismaeel, "Effective Technique for Allocating Servers to Support Cloud Using GPS and GIS," in Science and Information Conference 2013, London, 2013.

[29] Federal Communications Commission, 47CFR1.C.73.210 "Station Classes".

[30] United States Census Bureau, "2010 Census Interactive Population Search," 1 April 2010. [Online]. Available: http://www.census.gov/2010census/popmap/ipmtext.php?fl=12. [Accessed 1 April 2016].

[31] R. W. Stewart, K. W. Barlee, D. S. W. Atkinson and L. H. Crockett, Software Defined Radio Using MATLAB & Simulink and the RTL-SDR, Glasgow: Strathclyde Academic Media, 2015.

[32] Federal Communications Commission, 47CFR1.C.73.316 "FM antenna systems".

[33] MathWorks, "RTL-SDR Support from Matlab & Simulink," The MathWorks Inc., [Online]. Available: http://www.mathworks.com/hardware-support/rtl-sdr.html. [Accessed 28 May 2016].

[34] Federal Communications Commission, "FM Query Broadcast Station Search," 6 November 2014. [Online]. Available: http://www.fcc.gov/encyclopedia/fm- queary-broadcast-station-search.

[35] United States Geological Survey, "Domestic and Antartic Names - State and Topical Gazetteer Download Files," United States Geological Survey, 12 February 2016. [Online]. Available: http://geonames.usgs.gov/domestic/download_data.htm. [Accessed 4 April 2016].

[36] Wikipedia Foundation Inc., "Salt Lake City," Wikipedia Foundation Inc., 20 June 2016. [Online]. Available: https://en.wikipedia.org/wiki/Salt_Lake_City#Music. [Accessed 20 June 2016].

[37] Federal Communications Commission, 47CFR1.C.73.207 "Minimum distance separation between stations".

[38] Federal Communications Commission, "Digital Radio," Federal Communications Commission, 10 December 2015. [Online]. Available: https://www.fcc.gov/media/radio/digital-radio. [Accessed 29 June 2016].

[39] Federal Communications Commission, "Station Search Results," Federal Communications Commission, [Online]. Available: https://licensing.fcc.gov/cgi- bin/ws.exe/prod/cdbs/pubacc/prod/sta_list.pl. [Accessed 3 July 2016].

[40] G. V. Chaitanya, P. Rajalakshmi and U. B. Desai, "Real time hardware implementable spectrum sensor for Cognitive Radio applications," in 2012 International Conference on Signal Processing and Communications (SPCOM), Bangalore , 2012.

[41] Federal Communications Commission, 47CFR1.L.74.1201 "Definitions".

[42] Federal Communications Commission, 47CFR1.L.74.1235 "Power Limitations and Antenna Systems".

[43] United States Census Bureau, "Incorporated Places and Minor Civil Divisions Datasets: Subcounty Resident Population Estimates: April 1, 2010 to July 1, 2015," United States Census Bureau, 19 May 2016. [Online]. Available: http://www.census.gov/popest/data/cities/totals/2015/SUB-EST2015.html. [Accessed 18 June 2016].